Models

Claude 3.5 Sonnet balances intelligence and speed, and was the first Claude model to introduce computer use capabilities (screen navigation, clicking, typing) in public beta. It offered performance close to Claude 3 Opus at one-fifth the cost.

by Anthropic | $3/M input tokens | $15/M output tokens

Claude 3.7 Sonnet is Anthropic's first hybrid reasoning model, released February 2025. It combines instant responses with an extended thinking mode where users can control the "thinking budget" to balance speed vs. depth.

by Anthropic | $3/M input tokens | $15/M output tokens

Claude 3 Haiku is the fastest and most compact model from the Claude 3 family. It's optimized for near-instant responses and cost-efficiency, ideal for real-time chatbots, content moderation, and high-volume tasks.

by Anthropic | $0.25/M input tokens | $1.25/M output tokens

Claude Haiku 4.5 is Anthropic's small, fast model released October 2025, optimized for low latency and cost. Despite being the cheapest option ($1/$5 per million tokens), it matches Sonnet 4 on coding benchmarks (73.3% SWE-bench).

by Anthropic | $1/M input tokens | $5/M output tokens

Claude Opus 4.1 is an August 2025 incremental upgrade to Opus 4 focused on agentic tasks and real-world coding. It improved coding accuracy to 74.5% on SWE-bench with finer-grained refactoring and more precise bug fixes.

by Anthropic | $15/M input tokens | $75/M output tokens

Claude Opus 4 is the flagship model from the May 2025 Claude 4 launch, designed for complex long-running tasks. It can work continuously for several hours (7+ hour coding sessions demonstrated) and leads on coding benchmarks at 72.5% SWE-bench.

by Anthropic | $15/M input tokens | $75/M output tokens

Claude Opus 4.5 is Anthropic's most intelligent model, released November 2025. It sets the standard for production code, sophisticated agents, and complex enterprise tasks—scoring higher than Anthropic's own engineering candidates on technical tests.

by Anthropic | $5/M input tokens | $25/M output tokens

Claude Sonnet 4 is the May 2025 successor to Sonnet 3.7 with enhanced steerability and coding (72.7% SWE-bench). It excels at following complex instructions precisely and autonomous multi-feature app development with near-zero navigation errors.

by Anthropic | $3/M input tokens | $15/M output tokens

Claude Sonnet 4.5 is Anthropic's most capable model for agents and computer use, released September 2025. It can maintain focus for 30+ hours on complex tasks, supports a 1M token context window (beta), and is described as their "most aligned frontier model."

by Anthropic | $3/M input tokens | $15/M output tokens

DeepSeek Chat is the general-purpose conversational alias that points to the latest DeepSeek V3 chat model, a 671B parameter Mixture-of-Experts LLM optimized for everyday conversations, coding assistance, and general tasks. It supports 128K context and provides fast, direct responses without explicit reasoning chains.

by DeepSeek | $0.56/M input tokens | $1.68/M output tokens

DeepSeek Reasoner is the API alias for DeepSeek's reasoning models (R1 series), which use chain-of-thought reasoning to solve complex math, coding, and logic problems. It displays its thinking process before arriving at answers and achieves performance comparable to OpenAI o1.

by DeepSeek | $0.56/M input tokens | $1.68/M output tokens

Gemini 2.0 Flash is Google's fast multimodal model with native tool use, 1M token context window, and support for text, images, video, and audio input. It's optimized for agentic workflows with low latency and cost-efficient inference.

by Google | $0.1/M input tokens | $0.4/M output tokens

Gemini 2.0 Flash-Lite is Google's most cost-efficient model, optimized for large-scale text output tasks. It offers simplified pricing and lower costs than Flash while maintaining solid performance for high-volume workloads.

by Google | $0.08/M input tokens | $0.3/M output tokens

Gemini 2.5 Flash is Google's hybrid reasoning model balancing speed, cost, and intelligence with controllable thinking capabilities. It supports up to 1M tokens and excels at summarization, chat applications, and data extraction at scale.

by Google | $0.3/M input tokens | $2.5/M output tokens

Gemini 2.5 Flash-Lite is Google's cost-optimized version of 2.5 Flash, designed for high-volume tasks like classification, translation, and intelligent routing. It delivers efficient performance for cost-sensitive, high-scale operations.

by Google | $0.1/M input tokens | $0.4/M output tokens

Gemini 2.5 Pro is Google's most capable reasoning model with state-of-the-art performance on coding and complex tasks. It features a 1M token context window, advanced multimodal understanding, and Deep Think mode for enhanced reasoning.

by Google | $1.25/M input tokens | $10/M output tokens

Gemini 3 Flash is Google's frontier intelligence model built for speed, combining Pro-grade reasoning with Flash-level latency at a fraction of the cost. It excels at agentic coding, complex analysis, and multimodal understanding with configurable thinking levels.

by Google | $0.5/M input tokens | $3/M output tokens

Gemini 3 Pro is Google's most intelligent model, delivering state-of-the-art performance in reasoning, multimodal understanding, and agentic coding. It handles text, images, video, audio, and code with a 1M token context window and advanced tool-calling capabilities.

by Google | $2/M input tokens | $12/M output tokens

Codestral is Mistral's cutting-edge code generation model supporting 80+ programming languages with optimized low-latency performance. It specializes in fill-in-the-middle completion, code correction, and test generation with 2.5x faster performance than its predecessor.

by Mistral AI | $0.3/M input tokens | $0.9/M output tokens

Devstral Medium is a high-performance agentic coding model achieving 61.6% on SWE-Bench Verified. It excels at complex software engineering tasks across entire codebases, surpassing GPT-4.1 and Gemini 2.5 Pro in code-related tasks at a fraction of the cost.

by Mistral AI | $0.4/M input tokens | $2/M output tokens

Devstral Small is a 24B parameter agentic coding model built with All Hands AI, achieving 46.8% on SWE-Bench Verified. Released under Apache 2.0, it can run locally on a single RTX 4090 or 32GB RAM Mac for autonomous software development.

by Mistral AI | $0.1/M input tokens | $0.3/M output tokens

Magistral Medium is Mistral's enterprise reasoning model with chain-of-thought capabilities, scoring 73.6% on AIME2024 (90% with majority voting). It excels in multilingual step-by-step reasoning for legal, financial, and scientific applications.

by Mistral AI | $2/M input tokens | $5/M output tokens

Magistral Small is a 24B parameter open-source reasoning model under Apache 2.0, achieving 70.7% on AIME2024. It provides traceable, multilingual chain-of-thought reasoning in English, French, Spanish, German, Italian, Arabic, Russian, and Chinese.

by Mistral AI | $0.5/M input tokens | $1.5/M output tokens

Ministral 14B is part of the Ministral 3 family, a 14B parameter multimodal model with vision capabilities under Apache 2.0. It offers advanced capabilities for local deployment with instruct, base, and reasoning variants achieving 85% on AIME'25.

by Mistral AI | $0.2/M input tokens | $0.2/M output tokens

Ministral 3B is a compact 3B parameter multimodal model from the Ministral 3 family with vision capabilities. It runs on consumer hardware and edge devices, offering text and image understanding with 256K context in a 3-4GB quantized footprint.

by Mistral AI | $0.1/M input tokens | $0.1/M output tokens

Ministral 8B is an 8B parameter multimodal model offering best-in-class text and vision capabilities for edge deployment. It supports single-GPU operation and provides an optimal balance of performance and efficiency under Apache 2.0.

by Mistral AI | $0.15/M input tokens | $0.15/M output tokens

Mistral Large 3 is a 675B parameter sparse MoE model (41B active) trained on 3000 H200 GPUs, representing Mistral's frontier open-weight multimodal model. It supports 256K context, native vision, and excels in agentic workflows and enterprise applications.

by Mistral AI | $0.5/M input tokens | $1.5/M output tokens

Mistral Medium 3.1 is Mistral's frontier-class multimodal model released August 2025 with 128K context. It delivers near-frontier performance at $0.4/$2 per million tokens, excelling in reasoning, coding, and enterprise workflows.

by Mistral AI | $0.4/M input tokens | $2/M output tokens

Mistral Small 3.2 is a 24B parameter multimodal model with 128K context, improved instruction following, and reduced repetition. It handles text and images, runs on single RTX 4090 when quantized, and delivers 150 tokens/second under Apache 2.0.

by Mistral AI | $0.1/M input tokens | $0.3/M output tokens

Mistral 7B is Mistral's foundational 7.3B parameter open-source model under Apache 2.0, using sliding window attention and grouped-query attention. It outperforms Llama 2 13B on all benchmarks while being efficient enough for consumer hardware.

by Mistral AI | $0.25/M input tokens | $0.25/M output tokens

Mistral Nemo is a 12B parameter model built with NVIDIA featuring 128K context and the Tekken tokenizer trained on 100+ languages. It excels in multilingual tasks, coding, and reasoning, serving as a drop-in replacement for Mistral 7B.

by Mistral AI | $0.15/M input tokens | $0.15/M output tokens

Pixtral Large is a 124B parameter open-weights multimodal model built on Mistral Large 2, achieving frontier-level image understanding. It processes up to 30 high-resolution images per input with 128K context, excelling in document and chart analysis.

by Mistral AI | $2/M input tokens | $6/M output tokens

Voxtral Mini is a 3B parameter open-source speech model built on Ministral 3B under Apache 2.0. It handles transcription, Q&A from audio, and multilingual speech understanding for up to 40 minutes of audio, optimized for edge deployment.

by Mistral AI | $0.04/M input tokens | $0.04/M output tokens

Voxtral Small is a 24B parameter speech understanding model built on Mistral Small 3.1 under Apache 2.0. It supports 30-minute transcription, 40-minute audio understanding, Q&A, summarization, and function calling from voice in 8+ languages.

by Mistral AI | $0.1/M input tokens | $0.3/M output tokens

GPT-4.1 is OpenAI's smartest non-reasoning model, excelling at instruction following and tool calling with a 1M token context window. It outperforms GPT-4o across coding and multimodal tasks with a June 2024 knowledge cutoff.

by OpenAI | $2/M input tokens | $8/M output tokens

GPT-4.1 Mini is a smaller, faster version of GPT-4.1 that matches or exceeds GPT-4o performance while reducing latency by nearly half and cost by 83%. It features a 1M token context window and strong coding capabilities.

by OpenAI | $0.4/M input tokens | $1.6/M output tokens

GPT-4.1 Nano is OpenAI's fastest and cheapest model, designed for low-latency tasks like classification and autocompletion. It features a 1M token context window and scores 80.1% on MMLU despite its small size.

by OpenAI | $0.1/M input tokens | $0.4/M output tokens

GPT-4.5 Preview was OpenAI's largest pre-trained model focused on scaling unsupervised learning for improved creativity and reduced hallucinations. It has been deprecated in favor of GPT-4.1 and GPT-5 models.

by OpenAI | $75/M input tokens | $150/M output tokens

GPT-4o ("omni") is OpenAI's multimodal model capable of processing text, audio, images, and video inputs while generating text and images. It offers 4x faster responses than GPT-4 with superior non-English language and vision performance.

by OpenAI | $2.5/M input tokens | $10/M output tokens

GPT-4o Mini is a fast, affordable small model that scores 82% on MMLU and accepts text and image inputs. It's over 60% cheaper than GPT-3.5 Turbo while offering superior reasoning and coding capabilities.

by OpenAI | $0.15/M input tokens | $0.6/M output tokens

GPT-5 is OpenAI's unified reasoning system combining a fast model, a deeper thinking model, and an automatic router. It achieves 45% fewer factual errors than GPT-4o and sets state-of-the-art scores on math, coding, and health benchmarks.

by OpenAI | $1.25/M input tokens | $10/M output tokens

GPT-5 Chat is the non-reasoning version of GPT-5 used in ChatGPT, designed for conversational interactions. It's available as gpt-5-chat-latest in the API and provides fast responses without extended thinking.

by OpenAI | $1.25/M input tokens | $10/M output tokens

GPT-5 Mini is a faster, more cost-efficient version of GPT-5 optimized for well-defined tasks and precise prompts. It provides a balance between performance and speed for everyday use cases.

by OpenAI | $0.25/M input tokens | $2/M output tokens

GPT-5 Nano is OpenAI's fastest and cheapest GPT-5 variant, ideal for summarization and classification tasks. It offers extremely low latency for high-volume, simple inference workloads.

by OpenAI | $0.05/M input tokens | $0.4/M output tokens

GPT-5.1 is OpenAI's model that dynamically adapts reasoning time based on task complexity, making it faster and more token-efficient on simpler tasks. It features 8 customizable personalities and supports multimodal inputs.

by OpenAI | $1.25/M input tokens | $10/M output tokens

GPT-5.1 Chat is the conversational variant of GPT-5.1 used in ChatGPT with a warmer personality by default. It's available as gpt-5.1-chat-latest in the API for non-reasoning chat interactions.

by OpenAI | $1.25/M input tokens | $10/M output tokens

GPT-5.2 is OpenAI's flagship model for professional knowledge work and coding, outperforming industry professionals on GDPval across 44 occupations. It excels at spreadsheets, presentations, code, and complex multi-step projects.

by OpenAI | $1.75/M input tokens | $14/M output tokens

GPT-5.2 Chat is the ChatGPT-optimized variant of GPT-5.2 with an August 2025 knowledge cutoff. It provides conversational interactions with the latest world knowledge before requiring web search.

by OpenAI | $1.75/M input tokens | $14/M output tokens

OpenAI o1 is a reasoning model that thinks before answering using chain-of-thought, excelling at complex science and mathematics tasks. It was the first in OpenAI's "o" series designed for step-by-step logical reasoning.

by OpenAI | $15/M input tokens | $60/M output tokens

OpenAI o1 Mini was a faster, more affordable reasoning model alternative to o1, now deprecated in favor of o3-mini. It provided STEM-focused reasoning at lower cost and latency.

by OpenAI | $1.1/M input tokens | $4.4/M output tokens

OpenAI o1 Pro is a version of o1 with more compute for better responses, designed to think longer and provide the most reliable answers. It's the most expensive model at $150/1M input tokens.

by OpenAI | $150/M input tokens | $600/M output tokens

OpenAI o3 is a powerful reasoning model that pushes the frontier in coding, math, science, and visual perception. It can agentically use all ChatGPT tools and makes 20% fewer major errors than o1 on difficult tasks.

by OpenAI | $2/M input tokens | $8/M output tokens

OpenAI o3 Mini is a cost-efficient reasoning model specialized for STEM domains requiring precision and speed. It features three reasoning effort levels (low, medium, high) and supports function calling.

by OpenAI | $1.1/M input tokens | $4.4/M output tokens

OpenAI o4 Mini is a fast, cost-efficient reasoning model optimized for coding and visual tasks. It achieves remarkable performance for its size with full tool access in ChatGPT, succeeded by GPT-5 Mini.

by OpenAI | $1.1/M input tokens | $4.4/M output tokens

GPT-5.1 Codex is a version of GPT-5.1 optimized for agentic coding tasks in Codex or similar environments. It's designed for long-running coding workflows with enhanced code generation capabilities.

by OpenAI | $1.25/M input tokens | $10/M output tokens

GPT-5.1 Codex Mini is a smaller, more cost-effective version of GPT-5.1 Codex providing approximately 4x more usage within subscription limits. It balances coding capability with efficiency.

by OpenAI | $0.25/M input tokens | $2/M output tokens

GPT-5.2 Codex is OpenAI's most advanced agentic coding model for professional software engineering and defensive cybersecurity. It achieves state-of-the-art on SWE-Bench Pro with improved long-horizon work through context compaction.

by OpenAI | $1.75/M input tokens | $14/M output tokens

GPT-5.2 Pro is a version of GPT-5.2 that thinks longer to produce smarter and more precise responses for challenging problems. It supports medium, high, and xhigh reasoning effort settings.

by OpenAI | $21/M input tokens | $168/M output tokens

OpenAI o3 Pro is a version of o3 designed to think longer and provide the most reliable responses for challenging questions. It's recommended when reliability matters more than speed.

by OpenAI | $20/M input tokens | $80/M output tokens

Grok 2 is xAI's frontier language model released in August 2024, featuring advanced capabilities in chat, coding, and reasoning with competitive performance against GPT-4 and Claude 3.5 Sonnet. It integrates real-time information from the X platform and offers improved reasoning over Grok 1.5.

by xAI | $2/M input tokens | $10/M output tokens

Grok 2 Vision is a multimodal AI model that combines text and visual understanding capabilities, excelling at object recognition, visual math reasoning (MathVista), and document-based question answering (DocVQA). It supports image analysis with a 32K context window.

by xAI | $2/M input tokens | $10/M output tokens

Grok 3 is xAI's flagship model launched February 2025, trained with 10x more compute on the Colossus supercluster with 200,000 GPUs. It features advanced reasoning through reinforcement learning, deep domain knowledge in finance/healthcare/law/science, and a 131K token context window.

by xAI | $3/M input tokens | $15/M output tokens

Grok 3 Fast is a latency-optimized variant of Grok 3 using the same underlying model but served on faster infrastructure. It delivers quicker response times for latency-sensitive applications while maintaining equivalent reasoning quality and 131K context window.

by xAI | $5/M input tokens | $25/M output tokens

Grok 3 Mini is a lightweight, cost-efficient reasoning model that thinks before responding, ideal for logic-based tasks that don't require deep domain knowledge. It features configurable reasoning effort and exposes accessible thinking traces for transparency.

by xAI | $0.3/M input tokens | $0.5/M output tokens

Grok 3 Mini Fast is the speed-optimized variant of Grok 3 Mini, running on faster infrastructure for significantly quicker response times. It provides identical reasoning quality to Grok 3 Mini but is designed for latency-sensitive applications.

by xAI | $0.6/M input tokens | $4/M output tokens

Grok Beta was xAI's initial public API model released in late 2024, offering foundational chat and reasoning capabilities with 131K context window. It served as the enterprise API beta preview before being superseded by versioned Grok 2 models.

by xAI | $5/M input tokens | $15/M output tokens

Grok Vision Beta was the initial vision-enabled API model from xAI, providing image understanding and multimodal capabilities for processing text alongside visual inputs. It was released alongside grok-beta for enterprise API testing with an 8K context window.

by xAI | $5/M input tokens | $15/M output tokens

Jamba Large 1.7 is AI21's flagship open model with 94B active/398B total parameters, featuring a hybrid SSM-Transformer architecture with a 256K context window optimized for enterprise-grade reasoning tasks. It offers improved grounding and instruction-following capabilities, supporting multiple languages including English, Spanish, French, German, and Arabic.

by AI21 Labs | $2/M input tokens | $8/M output tokens

Jamba Mini 1.7 is AI21's lightweight and efficient open model built on the same hybrid SSM-Transformer architecture with a 256K context window, designed for fast and cost-effective everyday tasks. It delivers improved grounding and steerability in a compact form ideal for enterprise workflows like document analysis and RAG applications.

by AI21 Labs | $0.2/M input tokens | $0.4/M output tokens

Aion-1.0 is Aion Labs' most powerful reasoning model, a multi-model system built on DeepSeek-R1 and augmented with Tree of Thoughts (ToT) and Mixture of Experts (MoE) techniques. It excels at reasoning, coding, and various complex tasks with high benchmark performance.

by Aion Labs | $4/M input tokens | $8/M output tokens

Aion-1.0-Mini is a 32B parameter distilled version of DeepSeek-R1, designed for strong performance in mathematics, coding, and logic while being faster and more affordable. It outperforms R1-Distill-Qwen-32B and R1-Distill-Llama-70B on benchmarks.

by Aion Labs | $0.7/M input tokens | $1.4/M output tokens

Aion-RP-Llama-3.1-8B is an uncensored model fine-tuned for roleplaying and creative writing, ranking highest in the character evaluation portion of the RPBench-Auto benchmark. It supports 131K context and is optimized for immersive character chats, storytelling, and persona consistency.

by Aion Labs | $0.8/M input tokens | $1.6/M output tokens

CodeLlama 7B Instruct Solidity is a fine-tuned 7 billion parameter Code LLaMA model specialized for generating Solidity smart contracts. It was trained using 4-bit QLoRA fine-tuning on a dataset of 6,003 GPT-generated instruction and Solidity code pairs. The model takes human instructions and produces working Solidity code for blockchain applications.

Tongyi DeepResearch is an open-source agentic AI model from Alibaba with 30B total parameters (3B active per token) designed for deep, multi-step research tasks. It excels at autonomous web searching, browsing, and synthesizing information across sources with a 128K context window. The model achieves state-of-the-art results on research benchmarks like Humanity's Last Exam and BrowseComp.

by Alibaba | $0.09/M input tokens | $0.4/M output tokens

Molmo 2 8B is an open vision-language model from AI2 built on Qwen3-8B with SigLIP 2 vision backbone, supporting image, video, and multi-image understanding with state-of-the-art grounding and tracking capabilities. It outperforms larger models on video pointing benchmarks and excels at counting and captioning tasks.

by Allen AI | $0/M input tokens | $0/M output tokens

OLMo 2 32B Instruct (March 2025) is a fully open instruction-tuned language model from AI2, trained on the Dolma dataset with supervised fine-tuning, DPO, and RLVR. It's the first fully open model to outperform GPT-3.5 Turbo and GPT-4o mini on academic benchmarks while providing complete transparency into training data, code, and weights.

by Allen AI | $0.05/M input tokens | $0.2/M output tokens

OLMo 3 32B Think is AI2's flagship open reasoning model that uses long chain-of-thought thinking to improve performance on math and coding tasks. Trained on the Dolma 3 dataset with reinforcement learning from verifiable rewards, it's one of the strongest fully open reasoning models in its class.

by Allen AI | $0.15/M input tokens | $0.5/M output tokens

OLMo 3 7B Instruct is a lightweight fully open chat model from AI2, optimized for multi-turn dialogue, instruction-following, and tool use. Pre-trained on Dolma 3 and post-trained with SFT, DPO, and RLVR, it delivers strong performance while running on modest hardware.

by Allen AI | $0.1/M input tokens | $0.2/M output tokens

OLMo 3 7B Think is AI2's efficient reasoning model that surfaces intermediate thinking steps for complex prompts in math and coding. It matches Qwen 3 8B on MATH benchmarks and leads on HumanEvalPlus for coding while remaining fully open and accessible on consumer hardware.

by Allen AI | $0.12/M input tokens | $0.2/M output tokens

OLMo 3.1 32B Instruct is AI2's most capable fully open chat model, optimized for multi-turn conversation, instruction-following, and tool use. It represents the strongest fully open 32B-scale instruct model in AI2's evaluations, built on Dolma 3 with extended RL training.

by Allen AI | $0.2/M input tokens | $0.6/M output tokens

OLMo 3.1 32B Think is the updated flagship reasoning model from AI2 with an additional three weeks of reinforcement learning training beyond OLMo 3. It uses long chain-of-thought thinking for complex reasoning tasks and remains fully open with all training artifacts publicly available.

by Allen AI | $0.15/M input tokens | $0.5/M output tokens

Goliath 120B is a 120-billion parameter language model created by merging two fine-tuned Llama-2 70B models (Xwin and Euryale) using the mergekit framework. It excels at text generation, creative writing, question answering, and summarization tasks.

Amazon Nova 2 Lite is a fast, cost-effective multimodal reasoning model for everyday workloads that processes text, images, and video with a 1M token context window. It features extended thinking with adjustable reasoning intensity (low/medium/high) and built-in tools for web grounding and code execution. Released in December 2025, it excels at document processing, customer service chatbots, and agentic workflows.

by Amazon | $0.3/M input tokens | $2.5/M output tokens

Amazon Nova Lite is a very low-cost, lightning-fast multimodal model that processes text, images, and video inputs to generate text output. It supports up to 300K input tokens and can analyze multiple images or up to 30 minutes of video in a single request. Ideal for real-time customer interactions, document analysis, and visual question-answering tasks.

by Amazon | $0.06/M input tokens | $0.24/M output tokens

Amazon Nova Micro is a text-only model that delivers the lowest latency responses at the lowest cost in the Nova family. With a 128K token context window, it excels at text summarization, translation, content classification, interactive chat, and basic coding tasks. It's the fastest and most economical option when multimodal capabilities aren't needed.

by Amazon | $0.04/M input tokens | $0.14/M output tokens

Amazon Nova Premier is the most capable multimodal model in the Nova family, designed for complex reasoning tasks requiring the highest accuracy. It processes text, images, and video with advanced understanding capabilities and serves as the best teacher model for distilling custom variants of smaller Nova models. Best suited for sophisticated enterprise applications demanding top-tier intelligence.

by Amazon | $2.5/M input tokens | $12.5/M output tokens

Amazon Nova Pro is a highly capable multimodal model offering the best combination of accuracy, speed, and cost for a wide range of tasks. It supports up to 300K input tokens, excels at video summarization, financial document analysis, agentic workflows, and can process code bases with over 15,000 lines of code. It also serves as a teacher model for distilling custom variants of Nova Micro and Lite.

by Amazon | $0.8/M input tokens | $3.2/M output tokens

Magnum v4 72B is a 72-billion parameter language model fine-tuned on Qwen2.5-72B-Instruct, designed to replicate the prose quality of Claude 3 models (Sonnet and Opus). It excels at creative writing, roleplay, and conversational tasks with rich, contextually coherent text generation. The model supports up to 131K context length and multiple languages.

Claude 3.7 Sonnet (Thinking Mode) is Claude 3.7 Sonnet with extended thinking enabled by default. It excels at advanced math, competitive programming, and complex problem-solving by showing visible step-by-step reasoning.

by Anthropic | $3/M input tokens | $15/M output tokens

Arcee Coder Large is a 32B parameter code-specialized model based on Qwen 2.5-Instruct, fine-tuned on GitHub, CodeSearchNet, and synthetic bug-fix data for code generation and debugging. It supports 30+ programming languages with a 32k context window and shows 5-8 point gains over CodeLlama-34B-Python on HumanEval benchmarks. The model excels at producing compilable code with structured explanations, making it ideal for both educational tooling and production copilot scenarios.

by Arcee AI | $0.5/M input tokens | $0.8/M output tokens

Arcee Maestro Reasoning is a 32B parameter flagship analysis model derived from Qwen 2.5-32B, tuned with DPO and chain-of-thought reinforcement learning for step-by-step logical reasoning. It features a 128k context window and doubles pass-rates on MATH and GSM-8K benchmarks compared to its 7B predecessor. The model is designed for complex problem-solving, abstract reasoning, and scenario modeling with transparent reasoning traces suited for audit-focused industries.

by Arcee AI | $0.9/M input tokens | $3.3/M output tokens

Arcee Spotlight is a 7B parameter vision-language model derived from Qwen 2.5-VL, fine-tuned for image-text grounding tasks like captioning, visual question-answering, and diagram analysis. It offers a 32k-128k token context window enabling multimodal conversations combining documents with images. The model matches or outperforms larger VLMs like LLaVA-1.6 13B while running efficiently on consumer GPUs.

by Arcee AI | $0.18/M input tokens | $0.18/M output tokens

Arcee Trinity Mini is a 26B parameter sparse mixture-of-experts (MoE) model with only 3B active parameters per token, trained end-to-end in the U.S. on 10T tokens. It features 128 experts with 8 active per token, a 128k context window, and is optimized for multi-turn reasoning, function calling, and agent workflows. Released under Apache 2.0, it offers strong performance at extremely cost-efficient pricing.

by Arcee AI | $0.05/M input tokens | $0.15/M output tokens

Arcee Trinity Mini Free is the same 26B parameter MoE model (3B active) as Trinity Mini, offered through a free tier with rate limits on platforms like OpenRouter. It provides identical capabilities for multi-turn agent workflows, tool orchestration, and structured outputs with a 128k context window. This free version allows developers to experiment with the model before committing to paid usage.

by Arcee AI | $0/M input tokens | $0/M output tokens

Arcee Virtuoso Large is a 72B parameter general-purpose model based on Qwen 2.5-72B, trained using DistillKit and MergeKit with DeepSeek R1 distillation techniques. It retains a 128k context window for ingesting large documents, codebases, or financial filings, excelling at cross-domain reasoning, creative writing, and enterprise QA. The model serves as the fallback brain in Arcee Conductor pipelines when smaller SLMs flag low confidence.

by Arcee AI | $0.75/M input tokens | $1.2/M output tokens

ERNIE 4.5 21B A3B is a lightweight text-only language model from Baidu using a Mixture-of-Experts architecture with 21B total parameters but only 3B active per token. It excels at general language understanding, generation, reasoning, and coding tasks while remaining computationally efficient. Released under Apache 2.0, it achieves competitive performance against larger models like Qwen3-30B-A3B despite having 30% fewer total parameters.

by Baidu | $0.07/M input tokens | $0.28/M output tokens

ERNIE 4.5 21B A3B Thinking is Baidu's reasoning-enhanced language model built on the 21B A3B architecture with explicit chain-of-thought capabilities. It activates only 3B of its 21B parameters per token while specializing in logic, mathematics, coding, and multi-step reasoning tasks. The model supports extended context up to 131K tokens and is optimized for complex problem-solving through structured thinking.

by Baidu | $0.07/M input tokens | $0.28/M output tokens

ERNIE 4.5 300B A47B is Baidu's flagship text-only large language model featuring 300B total parameters with 47B active per token via MoE architecture. It demonstrates state-of-the-art performance on instruction following and knowledge benchmarks like IFEval, SimpleQA, and ChineseSimpleQA. The model supports 131K context length and excels at text understanding, generation, reasoning, and coding.

by Baidu | $0.28/M input tokens | $1.1/M output tokens

ERNIE 4.5 VL 28B A3B is a lightweight multimodal vision-language model with 28B total parameters but only 3B active per token. It processes both images and text simultaneously, enabling tasks like image comprehension, chart analysis, document understanding, and cross-modal reasoning. The model offers both thinking and non-thinking modes while matching performance of larger models like Qwen2.5-VL-32B.

by Baidu | $0.14/M input tokens | $0.56/M output tokens

ERNIE 4.5 VL 424B A47B is Baidu's largest multimodal vision-language model with 424B total parameters and 47B active per token. It supports up to 131K context tokens and excels at visual reasoning, document/chart understanding, and visual question answering with both thinking and non-thinking modes. In thinking mode, it approaches or surpasses OpenAI o1 on reasoning benchmarks like MathVista, MMMU, and VisualPuzzle.

by Baidu | $0.42/M input tokens | $1.25/M output tokens

Seed 1.6 is a general-purpose multimodal AI model by ByteDance featuring adaptive deep thinking, a 256K context window, and a sparse Mixture-of-Experts architecture with 230B total parameters (23B active per forward pass).

by ByteDance Seed | $0.25/M input tokens | $2/M output tokens

Seed 1.6 Flash is an ultra-fast multimodal model by ByteDance optimized for high-throughput and low-latency inference, supporting text, image, and video inputs with a 256K context window and up to 16K output tokens.

by ByteDance Seed | $0.08/M input tokens | $0.3/M output tokens

UI-TARS 1.5 7B is a multimodal vision-language agent by ByteDance optimized for GUI automation across desktop, web, mobile, and game environments. It uses reinforcement learning-based reasoning to plan and execute actions on graphical interfaces. The model achieves state-of-the-art results on benchmarks like OSWorld, WebVoyager, and AndroidWorld.

by ByteDance | $0.1/M input tokens | $0.2/M output tokens

Dolphin Mistral 24B Venice Edition is a 24-billion parameter uncensored AI model fine-tuned from Mistral-Small-24B, developed by Cognitive Computations in collaboration with Venice.ai. It's designed as a general-purpose assistant that gives users full control over system prompts and alignment, removing the typical safety filters found in mainstream models.

Command A is Cohere's most performant 111B parameter model with a 256K context length, excelling at tool use, RAG, agents, and multilingual tasks across 23 languages. It runs efficiently on just two GPUs while delivering 150% higher throughput than its predecessor Command R+. The model is optimized for enterprise applications including code generation, SQL, and complex reasoning tasks.

by Cohere | $2.5/M input tokens | $10/M output tokens

Command R 08-2024 is a 32B parameter multilingual model optimized for reasoning, summarization, and question answering with a 128K context length. It delivers 50% higher throughput and 20% lower latency than the original Command R while excelling at RAG, tool use, and code tasks. Trained on 23 languages, it features improved instruction following and Safety Modes.

by Cohere | $0.15/M input tokens | $0.6/M output tokens

Command R+ 08-2024 is a 104B parameter flagship model with a 128K context length, optimized for complex tasks including multi-step tool use and RAG workflows. It offers roughly 50% higher throughput and 25% lower latency compared to the previous Command R+ version. Trained on 23 languages, it excels at reasoning, summarization, and sophisticated enterprise applications.

by Cohere | $2.5/M input tokens | $10/M output tokens

Command R7B is a compact 7B parameter model released in December 2024, designed as the smallest and fastest in the Command R family with a 128K context length. It excels at RAG, tool use, and agentic tasks while being deployable on commodity hardware with minimal latency. The model supports 23 languages and is priced for high-volume enterprise applications.

by Cohere | $0.04/M input tokens | $0.15/M output tokens

Cogito v2 109B MoE is a sparse mixture-of-experts model built on Llama architecture that uses selective expert activation for efficient complex reasoning tasks. It features hybrid reasoning with both standard and self-reflective modes, achieving strong performance while being cost-effective at $0.18/$0.59 per million input/output tokens.

by Deep Cogito | $0.18/M input tokens | $0.59/M output tokens

Cogito v2 405B is a dense large-scale model built on Llama architecture, optimized for high performance across varied hardware setups. It offers hybrid reasoning capabilities and is designed to approach frontier-level intelligence within its size class through improved non-reasoning mode intuition.

by Deep Cogito | $3.5/M input tokens | $3.5/M output tokens

Cogito v2 70B is a dense model built on Llama architecture, ideal for low-latency applications and fine-tuning scenarios. It features dual-mode operation (standard LLM or self-reflective reasoning) and is optimized for environments with limited GPU capacity at $0.88 per million tokens.

by Deep Cogito | $0.88/M input tokens | $0.88/M output tokens

Cogito v2.1 671B is DeepCogito's flagship mixture-of-experts model with 671B total parameters (37B activated), ranking among the most powerful open-source AI models globally. It matches or exceeds DeepSeek R1 performance while using 60% shorter reasoning chains, supports 128k context and 30+ languages.

by Deep Cogito | $1.25/M input tokens | $1.25/M output tokens

DeepSeek V3-0324 is the March 2025 update to DeepSeek V3, incorporating reinforcement learning techniques from R1 to significantly improve reasoning, coding, and frontend development capabilities. It became the first open-source model to outperform all proprietary non-reasoning models on benchmarks, exceeding GPT-4.5 in math and coding tasks.

by DeepSeek | $0.19/M input tokens | $0.87/M output tokens

DeepSeek V3.1 is an August 2025 hybrid model that combines the capabilities of V3 and R1, supporting both thinking and non-thinking modes via chat template switching. It features 671B parameters (37B activated), 128K context, and significantly improved tool-calling and agent capabilities.

by DeepSeek | $0.15/M input tokens | $0.75/M output tokens

DeepSeek R1 is DeepSeek's first-generation reasoning model released January 2025, trained via large-scale reinforcement learning to achieve performance comparable to OpenAI o1 on math, code, and reasoning tasks. It pioneered open-source reasoning capabilities with self-verification and reflection behaviors.

by DeepSeek | $0.7/M input tokens | $2.5/M output tokens

DeepSeek R1-0528 is the May 2025 major update to R1, featuring dramatically improved reasoning depth with nearly double the thinking tokens (23K vs 12K average) and approaching performance of O3 and Gemini 2.5 Pro. It adds function calling support, reduced hallucinations, and improved AIME accuracy from 70% to 87.5%.

by DeepSeek | $0.4/M input tokens | $1.75/M output tokens

DeepSeek R1-0528 Free is the same May 2025 R1-0528 reasoning model offered through free API tiers on various providers, featuring the same improved reasoning capabilities and benchmark performance. It allows developers to test the model without cost, subject to rate limits.

by DeepSeek | $0/M input tokens | $0/M output tokens

DeepSeek R1 Distill Llama 70B is a 70 billion parameter dense model fine-tuned from Llama 3.3-70B-Instruct using 800K reasoning samples generated by DeepSeek R1. It brings R1's reasoning capabilities to a more accessible size while maintaining strong performance on math and coding benchmarks.

by DeepSeek | $0.03/M input tokens | $0.11/M output tokens

DeepSeek R1 Distill Qwen 32B is a 32 billion parameter dense model fine-tuned from Qwen 2.5 using R1-generated reasoning data, achieving state-of-the-art results for dense models. It outperforms OpenAI o1-mini on various benchmarks while being efficient enough for local deployment.

by DeepSeek | $0.29/M input tokens | $0.29/M output tokens

DeepSeek V3.1-Terminus is the September 2025 refined update to V3.1, addressing user-reported issues like language mixing and improving Code Agent and Search Agent capabilities. It represents the final, most stable version of the V3 architecture before V3.2.

by DeepSeek | $0.21/M input tokens | $0.79/M output tokens

DeepSeek V3.1-Terminus Exacto is a provider-specific variant of V3.1-Terminus, likely optimized for precise outputs or specific deployment configurations. It maintains the same hybrid thinking/non-thinking architecture with improved agentic tool use.

by DeepSeek | $0.21/M input tokens | $0.79/M output tokens

DeepSeek V3.2 is the December 2025 flagship model featuring DeepSeek Sparse Attention for efficiency and massive reinforcement learning post-training, achieving GPT-5-level performance. It's the first DeepSeek model to integrate thinking directly into tool-use and excels at agentic AI tasks.

by DeepSeek | $0.25/M input tokens | $0.38/M output tokens

DeepSeek V3.2-Exp is the September 2025 experimental predecessor to V3.2, introducing DeepSeek Sparse Attention architecture through continued training on V3.1-Terminus. It served as a testing ground for the sparse attention innovations later refined in V3.2.

by DeepSeek | $0.21/M input tokens | $0.32/M output tokens

DeepSeek V3.2-Speciale is a high-compute variant designed exclusively for maximum reasoning accuracy, achieving gold-medal performance in IMO 2025, IOI 2025, and ICPC World Finals. It rivals Gemini 3.0 Pro but requires higher token usage and doesn't support tool calling.

by DeepSeek | $0.27/M input tokens | $0.41/M output tokens

Llemma 7B is an open-source language model specialized for mathematics, built on Code Llama 7B and trained on 200B tokens of mathematical content. It excels at chain-of-thought mathematical reasoning and can use computational tools like Python and formal theorem provers. The model outperforms Llama-2, Code Llama, and Minerva when controlling for model size on math benchmarks.

Rnj-1 Instruct is an 8B parameter open-weight model by Essential AI, optimized for code and STEM tasks with state-of-the-art agentic coding capabilities. It excels at code generation, tool use, and mathematical problem-solving across multiple difficulty levels. The model is particularly strong on SWE-bench and supports 32K context length.

by Essential AI | $0.15/M input tokens | $0.15/M output tokens

Gemini 2.0 Flash 001 is a stable versioned release of Gemini 2.0 Flash, Google's fast multimodal workhorse model. It provides consistent behavior for production deployments with native tool use and 1M token context support.

by Google | $0.1/M input tokens | $0.4/M output tokens

Gemini 2.0 Flash Experimental (Free) is an experimental version of Gemini 2.0 Flash available at no cost with more restrictive rate limits. It's ideal for testing and development before moving to production with the stable release.

by Google | $0/M input tokens | $0/M output tokens

Gemini 2.0 Flash-Lite 001 is a stable versioned release of Google's most cost-efficient model. It's optimized for large-scale text tasks with simplified pricing and consistent behavior for production use.

by Google | $0.08/M input tokens | $0.3/M output tokens

Gemini 2.5 Flash-Lite Preview (September 2025) is a preview version of Google's cost-optimized Flash-Lite model. It's designed for high-volume classification, translation, and routing tasks with improved cost efficiency.

by Google | $0.1/M input tokens | $0.4/M output tokens

Gemini 2.5 Flash Preview (September 2025) is a preview version of Google's hybrid reasoning Flash model with controllable thinking capabilities. It balances quality, cost, and latency for enterprise-scale applications.

by Google | $0.3/M input tokens | $2.5/M output tokens

Gemini 2.5 Pro Preview is the preview version of Google's most advanced reasoning model with state-of-the-art coding and complex task performance. It features Deep Think mode, 1M token context, and advanced multimodal capabilities.

by Google | $1.25/M input tokens | $10/M output tokens

Gemini 2.5 Pro Preview (May 6) is a dated preview snapshot of Google's flagship reasoning model with improvements in code and function calling. It offers advanced reasoning capabilities for complex enterprise use cases.

by Google | $1.25/M input tokens | $10/M output tokens

Gemma 2 27B Instruct is Google's open-weight instruction-tuned language model with 27 billion parameters, trained on 13 trillion tokens. It offers competitive performance with models twice its size and runs on a single high-end GPU.

by Google | $0.65/M input tokens | $0.65/M output tokens

Gemma 2 9B Instruct is Google's efficient open-weight language model with 9 billion parameters, trained using knowledge distillation from the 27B model. It delivers strong performance for text generation while running on consumer hardware.

by Google | $0.03/M input tokens | $0.09/M output tokens

Gemma 3 12B Instruct is Google's mid-sized open multimodal model supporting text and image input with a 128K token context window. It supports 140+ languages and offers strong performance for single-GPU deployment.

by Google | $0.03/M input tokens | $0.1/M output tokens

Gemma 3 12B Instruct (Free) is the free-tier version of Google's 12B multimodal open model with image understanding and 128K context. It provides accessible AI capabilities for development and testing with rate limits.

by Google | $0/M input tokens | $0/M output tokens

Gemma 3 27B Instruct is Google's most capable single-GPU open model with multimodal support, 128K context, and 140+ language support. It outperforms many larger models and offers state-of-the-art open-weight performance.

by Google | $0.04/M input tokens | $0.15/M output tokens

Gemma 3 27B Instruct (Free) is the free-tier version of Google's flagship open model with multimodal capabilities and 128K context. It provides powerful AI for development with rate-limited free access.

by Google | $0/M input tokens | $0/M output tokens

Gemma 3 4B Instruct is Google's compact multimodal open model supporting text and images with a 128K token context window. It's optimized for deployment on laptops and edge devices while maintaining strong capabilities.

by Google | $0.02/M input tokens | $0.07/M output tokens

Gemma 3 4B Instruct (Free) is the free-tier version of Google's compact multimodal model suitable for edge deployment. It offers image understanding and 128K context for accessible development and testing.

by Google | $0/M input tokens | $0/M output tokens

Gemma 3n E2B Instruct (Free) is Google's mobile-first open model with an effective 2B parameter memory footprint using Per-Layer Embeddings. It's optimized for on-device AI with audio, text, image, and video understanding.

by Google | $0/M input tokens | $0/M output tokens

Gemma 3n E4B Instruct is Google's mobile-optimized model with a 4B active memory footprint containing a nested 2B submodel for flexible quality-latency tradeoffs. It supports real-time multimodal processing on edge devices.

by Google | $0.02/M input tokens | $0.04/M output tokens

Gemma 3n E4B Instruct (Free) is the free-tier version of Google's mobile-first 4B active parameter model with nested architecture. It enables on-device multimodal AI with audio, text, image, and video input capabilities.

by Google | $0/M input tokens | $0/M output tokens

MythoMax L2 13B is a 13-billion parameter model created by Gryphe that merges MythoLogic-L2 and Huginn using an experimental tensor merge technique. It excels at creative writing, roleplaying, and storytelling with enhanced coherency across narrative outputs. The model is based on the Llama 2 architecture and is available in various quantized formats for efficient deployment.

IBM Granite 4.0 Hybrid Micro is a 3B parameter dense hybrid model from IBM built on a decoder-only transformer architecture with Mamba2, optimized for low latency, edge, and local applications. It features enhanced tool-calling capabilities and is designed for fast execution of tasks like function calling within agentic workflows. Released under Apache 2.0 license, it's customizable via fine-tuning, LoRA adapters, and RAG.

Mercury is the world's first commercial diffusion large language model (dLLM) from Inception Labs that generates text 5-10x faster than traditional autoregressive LLMs by predicting multiple tokens in parallel. It's designed for latency-sensitive applications like voice agents, search interfaces, and chatbots while matching the quality of speed-optimized models like Claude 3.5 Haiku.

by Inception | $0.25/M input tokens | $1/M output tokens

Mercury Coder is a diffusion-based large language model specialized for code generation that achieves over 1,000 tokens per second on NVIDIA H100 GPUs. It's optimized for coding workflows including autocomplete, chat-based iteration, and code completion, delivering 5-10x faster speeds than models like GPT-4o Mini while maintaining comparable code quality.

by Inception | $0.25/M input tokens | $1/M output tokens

Inflection 3 Pi is an emotionally intelligent conversational AI model that powers Inflection's Pi chatbot, designed to adapt to your communication style and tone. It excels at customer support, roleplay, and empathetic conversations with access to recent news. The model features an 8K context window and prioritizes emotional intelligence and safety over task-oriented functions.

by Inflection AI | $2.5/M input tokens | $10/M output tokens

Inflection 3 Productivity is an enterprise-focused AI model optimized for precise instruction-following and structured output generation, particularly JSON. It features an 8K context window with access to recent news and is ideal for business automation, technical documentation, and workflow integration. Unlike its sibling Pi model, it prioritizes accuracy and compliance over emotional intelligence.

by Inflection AI | $2.5/M input tokens | $10/M output tokens

KAT-Coder-Pro V1 is KwaiKAT's most advanced agentic coding model, developed by Kuaishou's Kwaipilot team. It achieves a 73.4% solve rate on SWE-Bench Verified and is optimized for tool-use, multi-turn interaction, and real-world software engineering tasks through multi-stage training including agentic reinforcement learning.

by KwaiPilot | $0.21/M input tokens | $0.83/M output tokens

Liquid LFM2-2.6B is a 2.6 billion parameter hybrid language model from Liquid AI that combines grouped query attention with short convolutional layers for fast, efficient inference. It's optimized for on-device deployment on phones, laptops, and edge devices with strong multilingual support across 10 languages including English, Japanese, and Chinese.

by Liquid AI | $0.01/M input tokens | $0.02/M output tokens

Liquid LFM2-8B-A1B is Liquid AI's first on-device Mixture-of-Experts model with 8.3B total parameters but only 1.5B active per token, delivering 3-4B dense model quality at 1.5B-class compute. It runs faster than Qwen3-1.7B on mobile CPUs and is designed for private, low-latency applications on phones, tablets, and laptops.

by Liquid AI | $0.01/M input tokens | $0.02/M output tokens

Mancer Weaver (alpha) is a LLaMA 2 13B fine-tune designed for roleplay and narrative text generation, aiming to recreate Claude-style verbosity with an 8K context window. It's an unfiltered model offered through Mancer's API, optimized for creative storytelling scenarios rather than general-purpose tasks.

LongCat Flash Chat is a 560B parameter Mixture-of-Experts model by Meituan that dynamically activates only 18-31B parameters per token for efficiency. It supports 128K context length and excels at agentic tasks including multi-step reasoning, tool use, and coding. Released under MIT license in September 2025, it achieves over 100 tokens per second inference speed.

Llama 3 70B Instruct is a 70 billion parameter instruction-tuned language model from Meta, optimized for dialogue and assistant-like chat in English. It uses an optimized transformer architecture with grouped-query attention and was trained on over 15 trillion tokens.

by Meta Llama | $0.4/M input tokens | $0.4/M output tokens

Llama 3 8B Instruct is Meta's compact 8 billion parameter instruction-tuned model for dialogue use cases in English. It offers strong performance on common benchmarks while being more efficient to deploy than its larger sibling.

by Meta Llama | $0.03/M input tokens | $0.06/M output tokens

Llama 3.1 405B is Meta's flagship open-source large language model with 405 billion parameters, supporting 128K context length and 8 languages. It offers capabilities comparable to leading closed models for advanced reasoning, coding, and multilingual tasks.

by Meta Llama | $4/M input tokens | $4/M output tokens

Llama 3.1 405B Instruct is the instruction-tuned version of Meta's largest open model, optimized for multilingual dialogue, tool use, and complex reasoning. It supports 8 languages with 128K context and serves as a foundation for enterprise-level AI applications.

by Meta Llama | $3.5/M input tokens | $3.5/M output tokens

Llama 3.1 405B Instruct (Free) is a free-tier access version of Meta's 405B instruction-tuned model, offering the same multilingual dialogue and reasoning capabilities. It provides developers cost-free access to frontier-level open-source AI.

by Meta Llama | $0/M input tokens | $0/M output tokens

Llama 3.1 70B Instruct is a multilingual 70 billion parameter model with 128K context length, optimized for dialogue, tool use, and coding tasks. It balances strong performance with resource efficiency across 8 supported languages.

by Meta Llama | $0.4/M input tokens | $0.4/M output tokens

Llama 3.1 8B Instruct is Meta's efficient 8 billion parameter multilingual model supporting 128K context and 8 languages. It's ideal for resource-constrained deployments requiring summarization, classification, and translation capabilities.

by Meta Llama | $0.02/M input tokens | $0.05/M output tokens

Llama 3.2 11B Vision Instruct is Meta's multimodal model that processes both text and images with 11 billion parameters. It excels at visual recognition, image reasoning, captioning, and answering questions about images.

by Meta Llama | $0.05/M input tokens | $0.05/M output tokens

Llama 3.2 1B Instruct is Meta's ultra-lightweight 1 billion parameter model designed for edge and mobile devices. It supports 128K context and handles summarization, instruction following, and rewriting tasks locally.

by Meta Llama | $0.03/M input tokens | $0.2/M output tokens

Llama 3.2 3B Instruct is a compact 3 billion parameter model optimized for on-device use cases with 128K context support. It outperforms comparable models on instruction following, summarization, and tool-use tasks.

by Meta Llama | $0.02/M input tokens | $0.02/M output tokens

Llama 3.2 3B Instruct (Free) is a free-tier access version of Meta's 3 billion parameter edge-optimized model. It enables cost-free development of lightweight AI applications for mobile and edge devices.

by Meta Llama | $0/M input tokens | $0/M output tokens

Llama 3.3 70B Instruct is Meta's refined 70 billion parameter multilingual model with improved instruction following and tool use capabilities. It supports 8 languages and offers enhanced reasoning performance over previous versions.

by Meta Llama | $0.1/M input tokens | $0.32/M output tokens

Llama 3.3 70B Instruct (Free) is a free-tier access version of Meta's 70B model with enhanced multilingual dialogue capabilities. It provides developers cost-free access to strong reasoning and coding performance.

by Meta Llama | $0/M input tokens | $0/M output tokens

Llama 4 Maverick is Meta's 400 billion total parameter MoE model with 17B active parameters and 128 experts, supporting 1M token context. It's natively multimodal with state-of-the-art performance on coding, reasoning, and image understanding tasks.

by Meta Llama | $0.15/M input tokens | $0.6/M output tokens

Llama 4 Scout is Meta's efficient 109 billion parameter MoE model with 17B active parameters and 16 experts, featuring an industry-leading 10M token context window. It fits on a single H100 GPU and handles multimodal text and image inputs.

by Meta Llama | $0.08/M input tokens | $0.3/M output tokens

Llama Guard 2 8B is Meta's 8 billion parameter safety classifier built on Llama 3, designed to moderate both user prompts and AI responses. It classifies content across 11 hazard categories based on the MLCommons taxonomy.

by Meta Llama | $0.2/M input tokens | $0.2/M output tokens

Llama Guard 3 8B is Meta's enhanced safety moderation model providing content classification in 8 languages with support for tool call safety. It detects 14 hazard categories and integrates with Llama 3.1 for comprehensive AI safety.

by Meta Llama | $0.02/M input tokens | $0.06/M output tokens

Llama Guard 4 12B is Meta's 12 billion parameter multimodal safety model that moderates both text and image inputs across 12 languages. It was built from Llama 4 Scout and detects violations based on the MLCommons hazard taxonomy.

by Meta Llama | $0.18/M input tokens | $0.18/M output tokens

Phi-4 is a 14B parameter small language model from Microsoft that excels at complex reasoning tasks, especially mathematics, outperforming many larger models on math competition benchmarks while being efficient enough for edge deployment.

by Microsoft | $0.06/M input tokens | $0.14/M output tokens

WizardLM-2 8x22B is Microsoft AI's most advanced Wizard model, a Mixture of Experts LLM fine-tuned from Mixtral 8x22B that delivers near-GPT-4 performance on complex chat, multilingual, reasoning, and coding tasks while remaining open-source.

by Microsoft | $0.48/M input tokens | $0.48/M output tokens

MiniMax-01 is a 456B parameter foundation model (45.9B activated) using a hybrid Lightning Attention + MoE architecture, achieving top-tier performance on reasoning, math, and coding benchmarks. It supports up to 4 million tokens of context, making it especially strong for long-context tasks and AI agent applications.

by MiniMax | $0.2/M input tokens | $1.1/M output tokens

MiniMax-M1 is the world's first open-source hybrid-attention reasoning model, featuring a 1 million token context window and 80K reasoning output budget. It excels in software engineering, long-context tasks, and complex reasoning while being trained with an efficient CISPO reinforcement learning algorithm.

by MiniMax | $0.4/M input tokens | $2.2/M output tokens

MiniMax-M2 is a compact MoE model (230B total, 10B active parameters) optimized for coding and agentic workflows with a 128K context window. It ranks #1 among open-source models for tool use and agent tasks, delivering elite performance in multi-step development workflows at 8% the cost of comparable models.

by MiniMax | $0.2/M input tokens | $1/M output tokens

MiniMax-M2.1 is an enhanced version of M2 with significantly improved multi-language programming capabilities and office scenario support. It features more concise responses, better instruction following, and matches or exceeds Claude Sonnet 4.5 on coding benchmarks while maintaining excellent agent/tool scaffolding generalization.

by MiniMax | $0.27/M input tokens | $1.12/M output tokens

Devstral 2 is a 123B parameter dense transformer coding model achieving 72.2% on SWE-bench Verified with 256K context. Released under modified MIT license, it's the state-of-the-art open model for code agents, 7x more cost-efficient than Claude Sonnet.

by Mistral AI | $0.05/M input tokens | $0.22/M output tokens

Devstral 2 Free is the same 123B parameter frontier coding model available at no cost via API. It achieves 72.2% on SWE-bench Verified and excels at exploring codebases and orchestrating multi-file changes with architecture-level context.

by Mistral AI | $0/M input tokens | $0/M output tokens

Devstral Medium is a high-performance agentic coding model for complex software engineering tasks, achieving 61.6% on SWE-Bench Verified. It's designed for generalization across prompt styles and tool use in code agents and frameworks.

by Mistral AI | $0.4/M input tokens | $2/M output tokens

Devstral Small is a 24B parameter agentic LLM for software engineering, achieving 46.8% on SWE-Bench Verified. Released under Apache 2.0, it runs locally on consumer GPUs and excels at solving real-world GitHub issues autonomously.

by Mistral AI | $0.1/M input tokens | $0.3/M output tokens

Ministral 3B is a compact 3B parameter model optimized for edge deployment on phones, laptops, and IoT devices. It delivers robust multimodal capabilities in a small footprint, suitable for low-resource environments under Apache 2.0.

by Mistral AI | $0.04/M input tokens | $0.04/M output tokens

Ministral 8B is an 8B parameter model offering best-in-class text and vision capabilities for single-GPU operation. It provides an excellent balance of performance and efficiency for edge deployment and embedded applications.

by Mistral AI | $0.1/M input tokens | $0.1/M output tokens

Mistral 7B Instruct is the instruction-tuned version of Mistral 7B, fine-tuned on publicly available datasets. It outperforms all 7B models on MT-Bench and competes with 13B chat models while maintaining Apache 2.0 licensing.

by Mistral AI | $0.2/M input tokens | $0.2/M output tokens

Mistral 7B Instruct v0.1 is the original instruction-tuned version of Mistral 7B released September 2023. It demonstrates strong instruction-following capabilities while maintaining efficiency through sliding window and grouped-query attention.

by Mistral AI | $0.11/M input tokens | $0.19/M output tokens

Mistral 7B Instruct v0.2 introduces a 32K context window and improved performance over v0.1. It outperforms Llama 2 13B and Llama 1 34B on most benchmarks while remaining efficient for local deployment under Apache 2.0.

by Mistral AI | $0.2/M input tokens | $0.2/M output tokens

Mistral 7B Instruct v0.3 features an extended vocabulary with v3 Tokenizer and function calling support. It enhances language understanding and generation while maintaining the efficient 7B parameter architecture under Apache 2.0.

by Mistral AI | $0.2/M input tokens | $0.2/M output tokens

Mistral Large is Mistral's flagship large model for high-complexity enterprise tasks with strong reasoning, knowledge, and coding capabilities. It supports function calling and excels in RAG and agentic workflows across multiple languages.

by Mistral AI | $2/M input tokens | $6/M output tokens

Mistral Large 2 (24.07) is a 123B parameter model with 128K context, significantly upgraded for long context understanding and function calling. It delivers top-tier performance for enterprise use cases including knowledge exploration and automation.

by Mistral AI | $2/M input tokens | $6/M output tokens

Mistral Large 2 (24.11) includes improvements in long context understanding, system prompts, and function calling accuracy. Released alongside Pixtral Large, it's optimized for RAG and agentic workflows in enterprise deployments.

by Mistral AI | $2/M input tokens | $6/M output tokens

Mistral Medium 3 delivers frontier performance at $0.4/$2 per million tokens, performing at 90%+ of Claude Sonnet 3.7 across benchmarks. It's deployable on 4+ GPUs and surpasses Llama 4 Maverick and Cohere Command A.

by Mistral AI | $0.4/M input tokens | $2/M output tokens

Mistral Medium 3.1 (August 2025) is a frontier-class multimodal model with improved tone and performance. It features 128K context, native vision, and enhanced reasoning for STEM and enterprise workflows at competitive pricing.

by Mistral AI | $0.4/M input tokens | $2/M output tokens

Mistral Nemo is a 12B parameter model developed with NVIDIA featuring 128K context and the Tekken tokenizer. It's state-of-the-art in its class for reasoning, world knowledge, and coding in 11+ languages under Apache 2.0.

by Mistral AI | $0.02/M input tokens | $0.04/M output tokens

Mistral Saba is a 24B parameter regional model trained for Arabic and South Asian languages including Tamil and Malayalam. It outperforms models 5x its size on Arabic benchmarks while providing culturally relevant responses.

by Mistral AI | $0.2/M input tokens | $0.6/M output tokens

Mistral Small 3 is a 24B parameter latency-optimized model achieving ~81% MMLU accuracy at 150 tokens/second. It's designed for fast-response conversational agents and low-latency function calling under Apache 2.0.

by Mistral AI | $0.03/M input tokens | $0.11/M output tokens

Mistral Small 3.1 is a 24B multimodal model with 128K context, supporting text and image inputs. It outperforms GPT-4o Mini and Gemma 3 while delivering 150 tokens/second, released under Apache 2.0 for commercial use.

by Mistral AI | $0.03/M input tokens | $0.11/M output tokens

Mistral Small 3.1 Free provides the same 24B multimodal capabilities at no cost. It excels in vision understanding, multilingual tasks, and long context processing up to 128K tokens under Apache 2.0 licensing.

by Mistral AI | $0/M input tokens | $0/M output tokens

Mistral Small 3.2 improves on 3.1 with better instruction following (84.78% vs 82.75%), reduced infinite generations (1.29% vs 2.11%), and more robust function calling. It maintains the 24B/128K context architecture under Apache 2.0.

by Mistral AI | $0.06/M input tokens | $0.18/M output tokens

Mistral Small Creative is a specialized Labs model variant optimized for creative content generation. It builds on the Mistral Small architecture with adjustments for more imaginative and varied outputs in writing tasks.

by Mistral AI | $0.1/M input tokens | $0.3/M output tokens

Mistral Tiny is an earlier lightweight Mistral model optimized for speed and efficiency. It provides basic language capabilities for simple tasks where minimal latency and resource usage are prioritized over maximum performance.

by Mistral AI | $0.25/M input tokens | $0.25/M output tokens

Mixtral 8x22B is a sparse MoE model with 141B total / 39B active parameters, 64K context, and native function calling. It outperforms Llama 2 70B and matches GPT-3.5 while being cost-efficient under Apache 2.0.

by Mistral AI | $2/M input tokens | $6/M output tokens

Mixtral 8x7B is a sparse MoE model with 45B total / 13B active parameters using 8 experts per layer. It outperforms Llama 2 70B and GPT-3.5 while running 6x faster, mastering English, French, German, Spanish, and Italian.

by Mistral AI | $0.54/M input tokens | $0.54/M output tokens

Pixtral 12B is Mistral's first multimodal model with 12B text decoder + 400M vision encoder under Apache 2.0. It processes images at native resolution with 128K context, excelling in document QA and visual reasoning without compromising text performance.

by Mistral AI | $0.1/M input tokens | $0.1/M output tokens

Voxtral Small 24B is an open-source speech understanding model built on Mistral Small 3.1 under Apache 2.0. It handles transcription, translation, Q&A, and summarization directly from audio in 8+ languages with 32K token context.

by Mistral AI | $0.1/M input tokens | $0.3/M output tokens

Kimi Dev 72B is a 72-billion-parameter open-source coding LLM by Moonshot AI, built on Qwen 2.5-72B and optimized via reinforcement learning for software engineering tasks like bug fixing and code generation. It achieves state-of-the-art 60.4% on SWE-bench Verified among open-source models, using a BugFixer and TestWriter dual-component framework.

by Moonshot AI | $0.29/M input tokens | $1.15/M output tokens

Kimi K2 is Moonshot AI's flagship 1-trillion-parameter mixture-of-experts (MoE) model with 32 billion active parameters, optimized for agentic capabilities including tool use, reasoning, and coding. It achieves state-of-the-art performance among non-thinking models on knowledge, math, and coding benchmarks with 128K context support.

by Moonshot AI | $0.5/M input tokens | $2.4/M output tokens

Kimi K2 0905 is the September 2025 update to Kimi K2, extending context length from 128K to 256K tokens. It improves agentic coding accuracy, scaffold generalization, and frontend coding for web and 3D tasks while maintaining the same 1T parameter MoE architecture.

by Moonshot AI | $0.39/M input tokens | $1.9/M output tokens

Kimi K2 0905 Exacto is an OpenRouter routing variant that directs requests only to providers with verified high tool-calling accuracy and reliability. It uses the same Kimi K2 0905 model but filters for top-tier providers based on real-world telemetry, reducing tool-calling failures for agentic workflows.

by Moonshot AI | $0.6/M input tokens | $2.5/M output tokens

Kimi K2 Thinking is Moonshot AI's most advanced open-source reasoning model, built on the 1T parameter MoE architecture with step-by-step chain-of-thought reasoning and dynamic tool invocation. It sets state-of-the-art records on HLE (44.9%) and BrowseComp (60.2%), maintaining stable behavior across 200-300 sequential tool calls.

by Moonshot AI | $0.4/M input tokens | $1.75/M output tokens

Kimi K2 Free is a rate-limited free tier of Kimi K2 available through OpenRouter, providing no-cost access to the 1T parameter MoE model with restricted throughput. It's suitable for testing and prototyping but not recommended for production due to concurrency and rate limits.

by Moonshot AI | $0/M input tokens | $0/M output tokens

Morph V3 Fast is a specialized code-editing model optimized for speed, processing up to 10,500 tokens per second with 96% accuracy. It serves as an execution layer for AI-assisted development, applying code changes suggested by models like Claude or GPT-4o at unprecedented speeds.

by Morph | $0.8/M input tokens | $1.2/M output tokens

Morph V3 Large is a high-accuracy code transformation model processing at 4,500 tokens per second with 98% accuracy. It supports massive context windows up to 262,100 tokens and excels at applying code edits, refactoring, and systematic changes across entire codebases.

by Morph | $0.9/M input tokens | $1.9/M output tokens

Llama 3.1 Lumimaid 8B is a NeverSleep finetune of Meta's Llama 3.1 8B optimized for roleplay and creative writing with 32K context, balancing uncensored outputs with general conversational ability through 40% non-roleplay training data.

Noromaid 20B is a 20-billion parameter model by IkariDev and Undi designed for roleplay, ERP, and general conversation, trained on curated datasets including no_robots for human-like behavior and available in multiple quantized formats for flexible deployment.

DeepSeek V3.1 Nex N1 is a 670B parameter agentic AI model post-trained by Nex AGI for autonomous task execution, tool use, and real-world productivity. It excels at multi-step planning, code generation, web search integration, and complex reasoning tasks. The model is optimized for building and deploying AI agents with strong performance on programming, tool-use, and HTML generation benchmarks.

DeepHermes 3 Mistral 24B Preview is a 24B parameter instruction-tuned model based on Mistral-Small-24B, featuring a dual-mode system that toggles between intuitive chat responses and deep reasoning mode with extended chains of thought. It excels at function calling, structured JSON outputs, and multi-turn reasoning with the ability to use up to 13,000 tokens for complex problems.

by Nous Research | $0.02/M input tokens | $0.1/M output tokens

Hermes 2 Pro Llama 3 8B is an 8B parameter model fine-tuned on Meta's Llama 3, optimized for function calling (90% accuracy) and structured JSON outputs (84% accuracy). It features dedicated tool-call parsing tokens for agentic capabilities and outperforms Llama-3 8B Instruct on AGIEval, TruthfulQA, and BigBench benchmarks.

by Nous Research | $0.03/M input tokens | $0.08/M output tokens

Hermes 3 Llama 3.1 405B is a frontier-level 405B parameter full fine-tune of Llama-3.1-405B, focused on user alignment with powerful steering capabilities. It features advanced agentic capabilities, roleplaying, reasoning, multi-turn conversation, and improved code generation, competitive with or superior to Llama-3.1 Instruct models.

by Nous Research | $1/M input tokens | $1/M output tokens

Hermes 3 Llama 3.1 405B (Free) is the free-tier access version of the 405B parameter flagship Hermes 3 model. It offers the same frontier-level capabilities including advanced reasoning, roleplaying, function calling, and neutral alignment—just served through providers offering free API access.

by Nous Research | $0/M input tokens | $0/M output tokens

Hermes 3 Llama 3.1 70B is a 70B parameter fine-tune of Llama-3.1-70B offering advanced agentic capabilities, improved roleplaying, reasoning, and multi-turn conversation. It provides reliable function calling and structured outputs while being competitive with Llama-3.1 Instruct models at a more accessible size.

by Nous Research | $0.3/M input tokens | $0.3/M output tokens

Hermes 4 405B is a frontier hybrid-mode reasoning model based on Llama-3.1-405B, trained on a 60B token dataset with verified reasoning traces. It features toggleable deep reasoning via think tags, massive improvements in math, code, STEM, and logic, and achieves state-of-the-art on RefusalBench for reduced censorship.

by Nous Research | $1/M input tokens | $3/M output tokens

Hermes 4 70B is a hybrid reasoning model based on Llama-3.1-70B with toggleable deep thinking mode using think tags. It offers major improvements in math, code, STEM, logic, and creative writing while supporting JSON schema adherence, function calling, and reduced refusal rates compared to other models.

by Nous Research | $0.11/M input tokens | $0.38/M output tokens

Llama 3.1 Nemotron 70B Instruct is a 70B parameter LLM customized by NVIDIA using RLHF to improve response helpfulness, achieving top rankings on alignment benchmarks like Arena Hard and AlpacaEval 2 LC. It supports a 128K token context and is optimized for conversational AI and instruction-following tasks.

by NVIDIA | $1.2/M input tokens | $1.2/M output tokens

Llama 3.1 Nemotron Ultra 253B is a 253B parameter reasoning model derived from Llama 3.1 405B using Neural Architecture Search for improved efficiency, supporting 128K context and toggle ON/OFF reasoning modes. It excels at complex math, scientific reasoning, coding, RAG, and tool calling tasks while fitting on a single 8xH100 node.

by NVIDIA | $0.6/M input tokens | $1.8/M output tokens

Llama 3.3 Nemotron Super 49B v1.5 is an upgraded 49B parameter reasoning model derived from Llama 3.3 70B Instruct, optimized for single-GPU deployment on H100/H200 through Neural Architecture Search. It supports 128K context and is post-trained for agentic workflows including RAG, tool calling, and multi-turn conversations.

by NVIDIA | $0.1/M input tokens | $0.4/M output tokens

Nemotron 3 Nano 30B A3B is a 31.6B total parameter (3.2B active) hybrid Mamba-Transformer MoE model trained from scratch by NVIDIA with a 1M token context window. It offers up to 3.3x higher throughput than comparable models and supports configurable reasoning traces for both agentic and conversational tasks.

by NVIDIA | $0.06/M input tokens | $0.24/M output tokens

Nemotron 3 Nano 30B A3B (Free) is the free-tier access version of NVIDIA's hybrid Mamba-Transformer MoE reasoning model with 31.6B total parameters and only 3.2B active per token. It provides developers with efficient inference and customizable reasoning capabilities for building agentic AI systems at no cost.

by NVIDIA | $0/M input tokens | $0/M output tokens

Nemotron Nano 12B V2 VL is a 12.6B parameter multimodal vision-language model built on a hybrid Mamba-Transformer architecture for document intelligence and video understanding. It processes multiple images, documents, and videos while achieving leading results on OCRBench v2 with up to 2.5x higher throughput using Efficient Video Sampling.

by NVIDIA | $0.2/M input tokens | $0.6/M output tokens

Nemotron Nano 12B V2 VL (Free) is the free-tier version of NVIDIA's 12.6B multimodal vision-language model designed for document processing, visual QA, and video understanding. It combines transformer-level accuracy with efficient inference for tasks like invoice parsing, image summarization, and dense captioning.

by NVIDIA | $0/M input tokens | $0/M output tokens

Nemotron Nano 9B V2 is a 9B parameter hybrid Mamba-Transformer model trained from scratch by NVIDIA with a 128K context window, achieving up to 6x higher inference throughput than similar models like Qwen3-8B. It features controllable reasoning budget allowing developers to balance accuracy and response time for edge deployment.

by NVIDIA | $0.04/M input tokens | $0.16/M output tokens

Nemotron Nano 9B V2 (Free) is the free-tier access version of NVIDIA's efficient 9B hybrid Mamba-Transformer reasoning model with 128K context support. It provides controllable thinking budget and multilingual support across 10 languages for building AI agents, chatbots, and RAG systems.

by NVIDIA | $0/M input tokens | $0/M output tokens

ChatGPT-4o Latest is the GPT-4o model variant used in ChatGPT, not recommended for API use. It's optimized for ChatGPT's conversational interface rather than developer applications.

by OpenAI | $5/M input tokens | $15/M output tokens

GPT-3.5 Turbo is a legacy GPT model optimized for chat and non-chat tasks at low cost. As of July 2024, OpenAI recommends using GPT-4o Mini instead as it's cheaper, more capable, and multimodal.

by OpenAI | $0.5/M input tokens | $1.5/M output tokens

GPT-3.5 Turbo 0613 is a snapshot of GPT-3.5 Turbo from June 2023, providing consistent behavior for applications requiring a locked model version. It's a legacy model with limited support.

by OpenAI | $1/M input tokens | $2/M output tokens

GPT-3.5 Turbo 16K is a variant with an extended 16,384 token context window, allowing processing of longer documents. It's a legacy model superseded by newer models with larger contexts.

by OpenAI | $3/M input tokens | $4/M output tokens

GPT-3.5 Turbo Instruct is an instruction-following model using the Completions API rather than Chat Completions. It's designed for single-turn instruction tasks rather than multi-turn conversations.

by OpenAI | $1.5/M input tokens | $2/M output tokens

GPT-4 is an older high-intelligence GPT model that understands and generates complex text for creative writing, data analysis, and code generation. It has a 23,000-25,000 word context window.

by OpenAI | $30/M input tokens | $60/M output tokens

GPT-4 0314 is a snapshot of GPT-4 from March 2023, providing consistent behavior for applications requiring a specific model version. It's a legacy snapshot with limited ongoing support.

by OpenAI | $30/M input tokens | $60/M output tokens

GPT-4 1106 Preview is a November 2023 preview of GPT-4 Turbo with improved instruction following and JSON mode. It's a deprecated preview version superseded by GPT-4 Turbo's general release.

by OpenAI | $10/M input tokens | $30/M output tokens

GPT-4 Turbo is an older high-intelligence model designed as a cheaper, faster version of GPT-4 with a 128K context window. OpenAI now recommends using newer models like GPT-4o instead.

by OpenAI | $10/M input tokens | $30/M output tokens

GPT-4 Turbo Preview is a deprecated research preview of GPT-4 Turbo. It was an early fast GPT model that has been superseded by production releases.

by OpenAI | $10/M input tokens | $30/M output tokens

GPT-4o 2024-05-13 is the initial release snapshot of GPT-4o from May 2024. It provides version-locked behavior for applications requiring consistent model performance.

by OpenAI | $5/M input tokens | $15/M output tokens

GPT-4o 2024-08-06 is an August 2024 snapshot of GPT-4o with improvements and Structured Outputs support. It offers enhanced reliability for applications needing specific version behavior.

by OpenAI | $2.5/M input tokens | $10/M output tokens

GPT-4o 2024-11-20 is a November 2024 snapshot of GPT-4o providing the latest improvements at that time. It's useful for applications requiring locked model behavior.

by OpenAI | $2.5/M input tokens | $10/M output tokens

GPT-4o Audio Preview is a model for audio inputs and outputs with the Chat Completions API. It enables speech-in, speech-out conversational interactions and audio generation capabilities.

by OpenAI | $2.5/M input tokens | $10/M output tokens

GPT-4o Mini 2024-07-18 is the initial release snapshot of GPT-4o Mini from July 2024. It provides version-locked behavior for consistent performance in production applications.

by OpenAI | $0.15/M input tokens | $0.6/M output tokens

GPT-4o Mini Search Preview is a specialized model trained for web search queries in the Chat Completions API. It's a fast, affordable option for search-enabled applications.

by OpenAI | $0.15/M input tokens | $0.6/M output tokens

GPT-4o Search Preview is a specialized model for web search in Chat Completions, trained to understand and execute search queries. It returns responses with embedded citations and source references.

by OpenAI | $2.5/M input tokens | $10/M output tokens

GPT-4o Extended is a variant of GPT-4o with extended capabilities or context for specific use cases. It provides enhanced features beyond the standard GPT-4o model.

by OpenAI | $6/M input tokens | $18/M output tokens

GPT-5 Codex is a version of GPT-5 optimized for agentic coding tasks in Codex or similar environments. It's designed for software engineering workflows with enhanced code generation capabilities.

by OpenAI | $1.25/M input tokens | $10/M output tokens

GPT-5 Pro is an enhanced version of GPT-5 that thinks longer using parallel test-time compute to provide the highest quality answers. It replaces o3-pro for complex enterprise and research tasks.

by OpenAI | $15/M input tokens | $120/M output tokens

GPT-5.1 Codex Max is OpenAI's frontier agentic coding model built for long-running, detailed work using context compaction. It's the first model trained to operate across multiple context windows coherently.

by OpenAI | $1.25/M input tokens | $10/M output tokens

GPT-OSS 120B is OpenAI's most powerful open-weight model under Apache 2.0 license, achieving near-parity with o4-mini on reasoning benchmarks. It has 117B total parameters with 5.1B active, fitting on a single H100 GPU.

by OpenAI | $0.04/M input tokens | $0.19/M output tokens

GPT-OSS 120B Exacto is a variant of the open-weight model optimized for specific precision tasks. It maintains the Apache 2.0 license with full chain-of-thought access and configurable reasoning.

by OpenAI | $0.04/M input tokens | $0.19/M output tokens

GPT-OSS 120B Free is the freely accessible version of OpenAI's open-weight model through various providers. It offers the same capabilities as the standard model at no cost tier.

by OpenAI | $0/M input tokens | $0/M output tokens

GPT-OSS 20B is OpenAI's smaller open-weight model for lower latency and local inference, matching o3-mini on common benchmarks. It requires only 16GB of memory and runs on consumer hardware.

by OpenAI | $0.02/M input tokens | $0.1/M output tokens

GPT-OSS 20B Free is the freely accessible version of the smaller open-weight model through various inference providers. It's ideal for on-device use cases and rapid iteration without infrastructure costs.

by OpenAI | $0/M input tokens | $0/M output tokens

GPT-OSS Safeguard 20B is a safety-focused variant of the 20B open-weight model with additional content moderation capabilities. It includes enhanced safeguards for responsible deployment.

by OpenAI | $0.08/M input tokens | $0.3/M output tokens

OpenAI o3 Deep Research is a powerful model that searches and synthesizes hundreds of sources to create comprehensive research reports. It's optimized for browsing and data analysis at research analyst level.

by OpenAI | $10/M input tokens | $40/M output tokens

OpenAI o3 Mini High is a higher-intelligence version of o3-mini that takes longer to generate more accurate responses. It uses high reasoning effort for complex STEM and coding tasks.

by OpenAI | $1.1/M input tokens | $4.4/M output tokens

OpenAI o4 Mini Deep Research is a faster, more affordable deep research model for complex multi-step research tasks. It can synthesize information from web search and internal data sources.

by OpenAI | $2/M input tokens | $8/M output tokens

OpenAI o4 Mini High is a higher-intelligence version of o4-mini available in the ChatGPT model picker. It provides enhanced reasoning at the cost of longer response times.

by OpenAI | $1.1/M input tokens | $4.4/M output tokens

InternVL3 78B is an open-source multimodal large language model from OpenGVLab that combines a 6B vision encoder with a 72B language model to process images, videos, documents, and text. It achieves state-of-the-art performance among open-source MLLMs on both perception and reasoning benchmarks, rivaling closed-source models like GPT-4o. Key innovations include Variable Visual Position Encoding (V2PE), native multimodal pre-training, and Mixed Preference Optimization.

Sonar is Perplexity's lightweight, cost-effective search model built on Llama 3.3 70B, optimized for speed (1200 tokens/second) and quick factual queries. It provides real-time web search with grounding and citations, ideal for simple Q&A and straightforward integrations. Best for everyday use cases where fast, accurate answers are needed without complex reasoning.

by Perplexity | $1/M input tokens | $1/M output tokens

Sonar Deep Research is Perplexity's expert-level research model designed for exhaustive multi-step retrieval, synthesizing hundreds of sources into comprehensive reports. It autonomously searches, reads, and evaluates sources while refining its approach for in-depth analysis across domains like finance, technology, and health. Ideal for detailed market analyses, literature reviews, and projects requiring synthesis of multiple information sources.

by Perplexity | $2/M input tokens | $8/M output tokens

Sonar Pro is Perplexity's advanced search model with a 200K token context window, delivering 2x more citations and search results than standard Sonar for complex queries. It handles in-depth, multi-step queries with enhanced content understanding and supports longer, more nuanced follow-up conversations. Best for enterprise applications requiring deeper research and comprehensive source attribution.

by Perplexity | $3/M input tokens | $15/M output tokens

Sonar Pro Search is Perplexity's most advanced agentic search system, available exclusively via OpenRouter API, adding autonomous multi-step reasoning to Sonar Pro. Instead of single query+synthesis, it plans and executes entire research workflows using tools, making it ideal for deeper reasoning and analysis. This model powers Perplexity's Pro Search mode on their consumer platform.

by Perplexity | $3/M input tokens | $15/M output tokens

Sonar Reasoning Pro is a premier reasoning model powered by DeepSeek R1 with Chain of Thought (CoT), designed for complex multi-step analysis and logical problem-solving. It excels at tasks requiring step-by-step thinking, strict instruction adherence, and information synthesis across sources with built-in web search. Ranked among the top models in Search Arena evaluations, statistically tied with Gemini-2.5-Pro-Grounding.

by Perplexity | $2/M input tokens | $8/M output tokens

INTELLECT-3 is a 106B parameter open-source Mixture-of-Experts reasoning model (12B active parameters) from Prime Intellect, trained using large-scale reinforcement learning on top of GLM-4.5-Air-Base. It achieves state-of-the-art performance for its size across math, code, science, and reasoning benchmarks, scoring 90.8% on AIME 2024 and outperforming many larger frontier models.

by Prime Intellect | $0.2/M input tokens | $1.1/M output tokens

Qwen 2.5 72B Instruct is Alibaba's flagship open-source language model with 72 billion parameters, trained on 18 trillion tokens with 128K context support. It excels in coding, math, instruction following, and multilingual tasks across 29+ languages.

by Qwen | $0.12/M input tokens | $0.39/M output tokens

Qwen 2.5 7B Instruct is a compact yet capable language model offering strong performance in coding, math, and general tasks. It supports 128K context length and 29+ languages while being efficient enough for smaller deployments.

by Qwen | $0.04/M input tokens | $0.1/M output tokens

Qwen 2.5 Coder 32B Instruct is a code-specialized model matching GPT-4o's coding capabilities, supporting 40+ programming languages. It excels in code generation, repair, and reasoning with 128K context support.

by Qwen | $0.03/M input tokens | $0.11/M output tokens

Qwen 2.5 VL 7B Instruct is a vision-language model capable of understanding images, documents, charts, and videos up to 1 hour. It supports OCR, visual reasoning, and can act as a visual agent for computer/phone use.

by Qwen | $0.2/M input tokens | $0.2/M output tokens

Qwen 2.5 VL 7B Instruct (Free) is the free-tier version of the 7B vision-language model offering image/video understanding, OCR, and visual reasoning capabilities at no cost.

by Qwen | $0/M input tokens | $0/M output tokens

Qwen Max is Alibaba's most powerful proprietary API model, a large-scale MoE with hundreds of billions of parameters. It delivers top-tier performance in reasoning, coding, math, and multilingual tasks via Alibaba Cloud Model Studio.

by Qwen | $1.6/M input tokens | $6.4/M output tokens

Qwen Plus is a high-performance proprietary API model balancing capability and cost, suitable for complex tasks requiring strong reasoning and multilingual support. Available through Alibaba Cloud Model Studio.

by Qwen | $0.4/M input tokens | $1.2/M output tokens

Qwen Plus (2025-07-28) is a snapshot version of Qwen Plus from July 2025, offering consistent behavior and performance for production deployments requiring version stability.

by Qwen | $0.4/M input tokens | $1.2/M output tokens

Qwen Plus (2025-07-28) Thinking is the reasoning-enhanced version that uses chain-of-thought processing for complex problems, providing step-by-step reasoning before delivering answers.

by Qwen | $0.4/M input tokens | $4/M output tokens

Qwen Turbo is a fast, cost-effective API model with up to 1M context length, ideal for simple tasks requiring quick responses. It supports multiple languages and offers flexible tiered pricing.

by Qwen | $0.05/M input tokens | $0.2/M output tokens

Qwen VL Max is Alibaba's most capable vision-language API model based on Qwen2.5-VL, offering superior image/video understanding, OCR, document analysis, and visual reasoning capabilities.

by Qwen | $0.8/M input tokens | $3.2/M output tokens

Qwen VL Plus is a balanced vision-language API model offering good performance at lower cost, suitable for image understanding, OCR, and multimodal tasks without requiring maximum capability.

by Qwen | $0.21/M input tokens | $0.63/M output tokens

Qwen 2.5 Coder 7B Instruct is a compact code-specialized model with strong code generation, reasoning, and repair capabilities. It supports multiple programming languages while being deployable on consumer hardware.

by Qwen | $0.03/M input tokens | $0.09/M output tokens

Qwen 2.5 VL 32B Instruct is a mid-sized vision-language model offering enhanced image/video understanding with better alignment to human preferences. It bridges the gap between 7B and 72B variants.

by Qwen | $0.05/M input tokens | $0.22/M output tokens

Qwen 2.5 VL 72B Instruct is the flagship open-source vision-language model excelling in document understanding, visual reasoning, and long video comprehension up to 1 hour with event pinpointing.

by Qwen | $0.15/M input tokens | $0.6/M output tokens

Qwen3 14B is a dense language model with hybrid thinking/non-thinking modes, matching Qwen2.5-32B performance. It supports 119 languages and excels in math, coding, and reasoning tasks.

by Qwen | $0.05/M input tokens | $0.22/M output tokens

Qwen3 235B A22B is the flagship MoE model with 235B total and 22B active parameters, rivaling DeepSeek-R1 and o1. It features hybrid thinking modes and supports 119 languages with strong agentic capabilities.

by Qwen | $0.2/M input tokens | $0.6/M output tokens

Qwen3 235B A22B (2507) is the July 2025 updated version with significant improvements in instruction following, reasoning, coding, tool usage, and 256K long-context understanding.

by Qwen | $0.07/M input tokens | $0.46/M output tokens

Qwen3 235B A22B Thinking (2507) is the reasoning-enhanced variant using extended chain-of-thought processing for complex math, coding, and logical problems with enhanced performance.

by Qwen | $0.11/M input tokens | $0.6/M output tokens

Qwen3 30B A3B is an efficient MoE model with 30B total and 3B active parameters, outperforming QwQ-32B while using 10x fewer active parameters. It offers hybrid thinking modes and 119 language support.

by Qwen | $0.06/M input tokens | $0.22/M output tokens

Qwen3 30B A3B Instruct (2507) is the July 2025 updated instruction-tuned version with improved capabilities in reasoning, coding, and tool usage at high efficiency.

by Qwen | $0.08/M input tokens | $0.33/M output tokens

Qwen3 30B A3B Thinking (2507) is the reasoning-enhanced variant optimized for complex problem-solving with extended chain-of-thought processing at high parameter efficiency.

by Qwen | $0.05/M input tokens | $0.34/M output tokens

Qwen3 32B is a dense language model matching Qwen2.5-72B performance with hybrid thinking/non-thinking modes. It excels in STEM, coding, and reasoning while supporting 119 languages.

by Qwen | $0.08/M input tokens | $0.24/M output tokens

Qwen3 4B (Free) is a compact model rivaling Qwen2.5-72B-Instruct performance, featuring hybrid thinking modes and 119 language support. Available at no cost for lightweight deployments.

by Qwen | $0/M input tokens | $0/M output tokens

Qwen3 8B is a dense model matching Qwen2.5-14B performance with hybrid thinking modes and 128K context. It offers strong reasoning, coding, and multilingual capabilities in a mid-sized package.

by Qwen | $0.05/M input tokens | $0.25/M output tokens

Qwen3 Coder is the most agentic code model in the Qwen series, available in 30B and 480B MoE variants. It achieves SOTA on SWE-Bench with 256K native context, extendable to 1M tokens.

by Qwen | $0.22/M input tokens | $0.95/M output tokens

Qwen3 Coder 30B A3B Instruct is an efficient MoE coding model with 30B total and 3.3B active parameters, offering strong agentic coding capabilities with 256K context support.

by Qwen | $0.07/M input tokens | $0.27/M output tokens

Qwen3 Coder Flash is a cost-effective coding model balancing performance and speed, suitable for scenarios requiring fast responses at lower cost while maintaining coding quality.

by Qwen | $0.3/M input tokens | $1.5/M output tokens

Qwen3 Coder Plus is the strongest Qwen coding API model, ideal for complex project generation and in-depth code reviews with up to 1M token context support.

by Qwen | $1/M input tokens | $5/M output tokens

Qwen3 Coder Exacto is a precision-tuned variant optimized for accurate code generation with reduced errors, suitable for production code requiring high reliability.

by Qwen | $0.22/M input tokens | $1.8/M output tokens

Qwen3 Coder (Free) offers the flagship 480B MoE coding model at no cost, featuring SOTA agentic coding, function calling, and 256K context for repository-scale understanding.

by Qwen | $0/M input tokens | $0/M output tokens

Qwen3 Max is the most powerful Qwen3 API model with SOTA agent programming and tool usage capabilities. It features non-thinking mode optimized for complex agent scenarios.

by Qwen | $1.2/M input tokens | $6/M output tokens

Qwen3 Next 80B A3B Instruct is an innovative MoE model with hybrid attention (Gated DeltaNet + Gated Attention), achieving 10x inference throughput for 32K+ contexts while matching Qwen3-235B performance.

by Qwen | $0.09/M input tokens | $1.1/M output tokens

Qwen3 Next 80B A3B Instruct (Free) offers the innovative hybrid-attention architecture at no cost, providing efficient ultra-long context processing with 256K native support.

by Qwen | $0/M input tokens | $0/M output tokens

Qwen3 Next 80B A3B Thinking is the reasoning-enhanced variant outperforming Gemini-2.5-Flash-Thinking on complex reasoning tasks with hybrid attention and multi-token prediction.

by Qwen | $0.15/M input tokens | $1.2/M output tokens

Qwen3 VL 235B A22B Instruct is the flagship vision-language MoE model with 256K context, offering superior visual coding, spatial understanding, and long video comprehension up to 20 minutes.

by Qwen | $0.2/M input tokens | $1.2/M output tokens

Qwen3 VL 235B A22B Thinking is the reasoning-enhanced vision-language model excelling at visual math, detail analysis, and causal reasoning with extended chain-of-thought processing.

by Qwen | $0.98/M input tokens | $3.95/M output tokens

Qwen3 VL 30B A3B Instruct is an efficient vision-language MoE model offering strong image/video understanding with 3B active parameters and 256K context support.

by Qwen | $0.15/M input tokens | $0.6/M output tokens

Qwen3 VL 30B A3B Thinking is the reasoning-enhanced vision-language variant optimized for complex visual reasoning tasks with extended thinking capabilities.

by Qwen | $0.2/M input tokens | $1/M output tokens

Qwen3 VL 32B Instruct is a dense vision-language model with strong text and visual capabilities, featuring visual coding, spatial understanding, and 256K context support.

by Qwen | $0.5/M input tokens | $1.5/M output tokens

Qwen3 VL 8B Instruct is a compact vision-language model matching flagship text performance while supporting image/video understanding, visual coding, and 256K context length.

by Qwen | $0.08/M input tokens | $0.5/M output tokens

Qwen3 VL 8B Thinking is the reasoning-enhanced compact vision model for complex visual analysis requiring step-by-step reasoning with efficient resource usage.

by Qwen | $0.18/M input tokens | $2.1/M output tokens

QwQ 32B is a 32B parameter reasoning model rivaling DeepSeek-R1 (671B) through scaled reinforcement learning. It excels in math, coding, and complex reasoning with 131K context and agent capabilities.

by Qwen | $0.15/M input tokens | $0.4/M output tokens

SorcererLM 8x22B is an advanced roleplay and storytelling model built as a LoRA fine-tune on WizardLM-2 8x22B, designed to improve vocabulary and writing style for creative fiction. It features enhanced emotional intelligence, vivid writing with spatial awareness, and dynamic narrative depth for immersive interactions.

Relace Apply 3 is a specialized code-patching model that merges AI-generated code edits into existing source files at up to 10,000 tokens per second. It supports a 256K context window and works with diffs from models like Claude and GPT-4, making code integration fast and reliable.

by Relace | $0.85/M input tokens | $1.25/M output tokens

Relace Search is an agentic codebase search model that uses 4-12 parallel tool calls (view_file, grep) to explore repositories and return relevant files. It performs multi-step reasoning to produce precise results 4x faster than frontier models, designed to work as a subagent for coding workflows.

by Relace | $1/M input tokens | $3/M output tokens

Llama 3 Euryale 70B v2.1 is a 70-billion parameter model by Sao10K focused on creative roleplay and storytelling, featuring strong prompt adherence, spatial awareness, and non-restrictive creative writing capabilities. It adapts well to custom formatting and produces highly varied, creative outputs.

Llama 3 8B Lunaris is a versatile 8B parameter generalist and roleplaying model created by merging five different Llama 3-based models. It balances creativity with improved logical reasoning and general knowledge, serving as an evolution of Stheno v3.2.

Llama 3.1 70B Hanami x1 is an experimental 70B model built on top of Euryale v2.2 by Sao10K, offering a different feel with enhanced creativity and logical reasoning. The creator considers it an improvement over both Euryale v2.1 and v2.2.

Llama 3.1 Euryale 70B v2.2 is Sao10K's creative roleplay model built on Meta's Llama 3.1 architecture with improved multi-turn coherency, system prompt handling, and reasoning capabilities. It features a 32K context window and excels at immersive storytelling with strong prompt adherence.

Llama 3.3 Euryale 70B v2.3 is the latest in Sao10K's Euryale series, built on Llama 3.3 Instruct with a 131K context window and 16K output limit. It's a direct successor to v2.2, trained without LoRA extraction for more robust creative roleplay and storywriting performance.

Step3 is a multimodal reasoning model from StepFun built on a Mixture-of-Experts architecture with 321B total parameters (38B active), designed for cost-effective vision-language reasoning. It features innovative Multi-Matrix Factorization Attention (MFA) and Attention-FFN Disaggregation (AFD) for efficient inference across various hardware. The model excels at visual interpretation, complex reasoning, and is available via API or as open-source checkpoints on Hugging Face.

Switchpoint Router is an intelligent LLM routing system that automatically analyzes your request and directs it to the optimal AI model from its library. It uses a cascading approach, trying lower-cost models first and escalating only when needed, optimizing for cost-performance balance. As new models emerge, the router adapts automatically without requiring changes to your workflow.

by SwitchPoint | $0.85/M input tokens | $3.4/M output tokens

Hunyuan A13B Instruct is an open-source large language model from Tencent that uses a Mixture-of-Experts architecture with 80B total parameters but only 13B active during inference, making it highly efficient. It supports dual reasoning modes (fast and slow thinking), a 256K context window, and excels at mathematics, science, coding, and agent tasks.

by Tencent | $0.14/M input tokens | $0.57/M output tokens

Cydonia 24B v4.1 is a 24-billion parameter uncensored creative writing model based on Mistral Small 3.2, optimized for roleplay, storytelling, and long-form narratives with a 131K token context window. It excels at character consistency, descriptive prose without being overly flowery, and maintains good recall and prompt adherence. The model also performs well for coding and instruction-following tasks.

Rocinante 12B is a 12-billion parameter creative writing model built on the Mistral architecture, designed for adventure-filled storytelling, roleplay, and imaginative text generation. Named after Don Quixote's horse, it produces rich, distinct prose with enhanced vocabulary and supports multiple chat templates including ChatML, Alpaca, and Mistral. The model offers a good balance between creative capability and computational efficiency for local deployment.

Skyfall 36B v2 is a 36-billion parameter model that upscales Mistral Small 2501 with specialized training for creativity, roleplay, and coherent storytelling. Users report it rivals or exceeds 70B parameter models in creative writing quality while remaining accessible for local deployment with strong chain-of-thought reasoning and tool use capabilities. It features a 32K token context window and supports Mistral v7 Tekken, Metharme, and Alpaca chat templates.

UnslopNemo 12B is a 12-billion parameter model where TheDrummer removed repetitive patterns ('slop') from roughly 90% of the roleplay training dataset to make outputs more expressive and varied. It's designed for adventure writing and roleplay scenarios with a 32K token context window. The model aims to generate more natural, less formulaic creative content compared to standard fine-tuned models.

DeepSeek R1T Chimera is a 685B parameter model created by TNG Tech that merges DeepSeek-R1's reasoning capabilities with DeepSeek-V3's token efficiency. It uses 40% fewer output tokens than R1 while maintaining similar intelligence, constructed via a novel Assembly-of-Experts method rather than fine-tuning.

DeepSeek R1T Chimera (Free) is the free-tier version of TNG Tech's merged model combining DeepSeek-R1 and DeepSeek-V3. It offers R1-level reasoning with improved token efficiency, using 40% fewer output tokens while being constructed from neural network parts of both parent MoE models.

DeepSeek R1T2 Chimera is TNG Tech's second-generation 671B parameter tri-parent model assembled from DeepSeek R1-0528, R1, and V3-0324. It runs ~20% faster than R1 and 2x faster than R1-0528 while scoring higher on benchmarks like GPQA and AIME-24, with improved think-token consistency.

DeepSeek R1T2 Chimera (Free) is the free-tier version of TNG Tech's second-generation tri-parent Chimera model. It combines three DeepSeek models via Assembly-of-Experts merge, offering 2x faster inference than R1-0528 with strong reasoning performance and up to 130K context support.

TNG R1T Chimera is an experimental LLM from TNG Tech optimized for creative storytelling and character interaction. It's a derivative of the original DeepSeek-R1T-Chimera with improved think-token consistency, better tool calling, and an EQ-Bench3 score of ~1305.

TNG R1T Chimera (Free) is the free-tier version of TNG Tech's creative-focused experimental model. It specializes in storytelling and character interaction with a pleasant personality, featuring improved tool calling and more consistent reasoning blocks compared to the original R1T.

ReMM SLERP L2 13B is a 13 billion parameter language model created by Undi95 as a recreation of MythoMax-L2-13B, using SLERP merging to combine ReML and Huginn v1.2 models. It's built on Llama 2 and merges several fine-tuned models including Chronos-Beluga, Airoboros, and Nous-Hermes for enhanced text generation capabilities.

Grok 3 Beta is the API-accessible version of Grok 3, xAI's most advanced model with superior reasoning, mathematics, coding, and world knowledge capabilities refined through large-scale reinforcement learning. It supports enterprise data extraction, coding, and text summarization tasks.

by xAI | $3/M input tokens | $15/M output tokens

Grok 3 Mini Beta is the API version of Grok 3 Mini, a cost-efficient lightweight reasoning model with configurable reasoning effort (low/high) parameters. It excels at STEM tasks and logic-based problems while exposing transparent thinking traces.

by xAI | $0.3/M input tokens | $0.5/M output tokens

Grok 4 is xAI's flagship reasoning model released July 2025, trained with unprecedented reinforcement learning scale on 200,000 GPUs. It features native tool use, real-time search integration, and Grok 4 Heavy achieves 50% on Humanity's Last Exam benchmark.

by xAI | $3/M input tokens | $15/M output tokens

Grok 4 Fast is an optimized variant delivering performance similar to Grok 4 but using 40% fewer thinking tokens with a massive 2 million token context window. It offers up to 64x cost reduction versus early frontier models like OpenAI's o3.

by xAI | $0.2/M input tokens | $0.5/M output tokens

Grok 4.1 Fast is xAI's best tool-calling model released November 2025, featuring a 2M context window and halved hallucination rates versus Grok 4 Fast. It's optimized for agentic workflows with native support for web search, X search, and code execution.

by xAI | $0.2/M input tokens | $0.5/M output tokens

Grok Code Fast 1 is a speedy, economical reasoning model built from scratch specifically for agentic coding workflows, released August 2025. It excels at TypeScript, Python, Java, Rust, C++, and Go with a 256K context window and ~92 tokens/second throughput.

by xAI | $0.2/M input tokens | $1.5/M output tokens

MiMo-V2-Flash is Xiaomi's open-source Mixture-of-Experts language model with 309B total parameters (15B active), designed for high-speed reasoning, coding, and agentic workflows. It uses a hybrid attention architecture with Multi-Token Prediction to achieve up to 150 tokens/second inference while keeping costs extremely low. The model excels at software engineering benchmarks and supports a 256K context window.

by Xiaomi | $0.09/M input tokens | $0.29/M output tokens

MiMo-V2-Flash (Free) is the free-tier version of Xiaomi's MiMo-V2-Flash model available through API providers like OpenRouter. It offers the same 309B parameter MoE architecture with 15B active parameters, hybrid-thinking toggle, and 256K context window as the paid version. It ranks as the top open-source model on SWE-bench benchmarks while being offered at no cost for users to try.

by Xiaomi | $0/M input tokens | $0/M output tokens

GLM-4-32B is a 32-billion parameter bilingual (Chinese-English) foundation model by Zhipu AI, pre-trained on 15TB of reasoning-focused data. It delivers performance comparable to GPT-4o on code generation, function calling, and Q&A tasks while remaining deployable on accessible hardware.

by Z.AI | $0.1/M input tokens | $0.1/M output tokens

GLM-4.5 is Zhipu AI's flagship 355B-parameter open-source model (32B active) designed for agentic AI applications with dual thinking/non-thinking modes. It excels at reasoning, coding, and tool use, ranking 3rd globally among all models on combined benchmarks under MIT license.

by Z.AI | $0.35/M input tokens | $1.55/M output tokens

GLM-4.5-Air is a compact 106B-parameter variant (12B active) of GLM-4.5, offering competitive agentic performance with significantly lower resource requirements. It supports the same dual reasoning modes and 128K context window as its larger sibling.

by Z.AI | $0.05/M input tokens | $0.22/M output tokens

GLM-4.5-Air Free is the same efficient 106B/12B-active model as GLM-4.5-Air, offered at no cost for developers to experiment with agentic workflows, coding, and reasoning tasks.

by Z.AI | $0/M input tokens | $0/M output tokens

GLM-4.5V is a 106B-parameter vision-language model achieving SOTA on 42 multimodal benchmarks, capable of image/video reasoning, GUI agent tasks, document parsing, and visual grounding. It features a thinking mode toggle and 64K multimodal context under MIT license.

by Z.AI | $0.6/M input tokens | $1.8/M output tokens

GLM-4.6 is Zhipu AI's 355B-parameter (32B active) flagship text model with 200K context, excelling at coding, agentic workflows, and search tasks. It's 15% more token-efficient than GLM-4.5 and ranks as the #1 domestic model in China.

by Z.AI | $0.35/M input tokens | $1.5/M output tokens

GLM-4.6 Exacto is a precision-optimized variant of the GLM-4.6 model, likely tuned for higher accuracy on specific tasks requiring exact outputs. It shares the same 355B MoE architecture with enhanced performance on structured generation.

by Z.AI | $0.44/M input tokens | $1.76/M output tokens

GLM-4.6V is a 106B vision-language model featuring native multimodal Function Calling—the first to directly pass images as tool inputs. It supports 128K context for processing 150+ page documents or 1-hour videos in a single pass.

by Z.AI | $0.3/M input tokens | $0.9/M output tokens

GLM-4.7 is Zhipu AI's latest ~400B flagship released December 2025, optimized for coding with 200K context and 128K output. It scores 73.8% on SWE-bench and 95.7% on AIME 2025, featuring Preserved Thinking for consistent multi-turn reasoning at $3/month.

by Z.AI | $0.4/M input tokens | $1.5/M output tokens