Xiaomi MiMo API

MiMo-V2.6-Flash is Xiaomi's cost-efficient omnimodal model in the MiMo-V2.6 series, a sparse Mixture-of-Experts model with 309B total parameters (15B active) that processes text, image, video, and audio in a single architecture with a 1M-token context window. It uses a hybrid attention design that interleaves sliding-window and global attention with a lightweight Multi-Token Prediction module, and scores 87.6 on Terminal-Bench 2.1 and 95.1 on CyberGym, staying close to its larger sibling MiMo-V2.6-Pro on several agent benchmarks despite the lower price. Priced at $0.14 per million input tokens and $0.28 per million output tokens, MiMo-V2.6-Flash fits agent pipelines that need multimodal input and long context without paying flagship rates.

MiMo-V2.6-Pro is Xiaomi's flagship model in the MiMo-V2.6 series, a sparse Mixture-of-Experts model with 1.02 trillion total parameters (42B active) and a 1M-token context window, built for long-horizon agentic work across coding, tool use, and multimodal reasoning. It scored 46 on the Artificial Analysis Intelligence Index, the highest reported for an open-weight model at release, and reaches 89.9 on Terminal-Bench 2.1 and 94.0 on CyberGym. Artificial Analysis measured its cost per index task at $0.13, against $5.86 for Claude Opus 5. Priced at $0.44 per million input tokens and $0.87 per million output tokens, MiMo-V2.6-Pro targets developers who want agentic performance close to frontier proprietary models at a fraction of the cost.

MiMo-V2.6-Pro-UltraSpeed is a latency-optimized serving variant of MiMo-V2.6-Pro, built from the same 1.02-trillion-parameter checkpoint and the same 1M-token context window, tuned for response speed rather than lowest cost per token. Xiaomi describes it as running MiMo-V2.6-Pro at up to 20x faster output, with early third-party measurements putting the gain closer to 10x throughput. Because it shares Pro's weights, its accuracy and benchmark scores track the base model; only the serving path changes. Priced at $4.35 per million input tokens and $8.70 per million output tokens, roughly ten times standard Pro pricing, MiMo-V2.6-Pro-UltraSpeed is meant for latency-sensitive agent and chat applications, not cost-sensitive batch workloads.

MiMo V2.5 is a native omnimodal model from Xiaomi that processes text, images, video, and audio within a single architecture and a 1M-token context window. It delivers agentic performance close to its larger sibling, MiMo V2.5 Pro, at roughly half the token cost — scoring 62.3 on ClawEval (general) and 23.8 on ClawEval Multimodal. On video understanding, it reaches 87.7 on Video-MME, competitive with Gemini 3 Pro. Image understanding benchmarks include 81.0 on CharXiv RQ and 77.9 on MMMU-Pro. Priced at $0.40 per million input tokens and $2.00 per million output tokens, MiMo V2.5 is a strong fit for production agent pipelines where you need multimodal perception and reasoning without flagship-tier cost.

MiMo V2.5 Pro is Xiaomi's most capable model, built for complex software engineering, long-horizon agentic tasks, and autonomous multi-step workflows spanning over a thousand tool calls. It scores 57.2 on SWE-bench Pro, 63.8 on ClawEval, and 72.9 on τ3-Bench — placing it alongside Claude Opus 4.6 and GPT-5.4 across most agentic evaluations. Notably, it achieves this while using roughly 40–60% fewer tokens per trajectory than comparable frontier models. The 1M-token context window and 131K max output support entire codebases and extended autonomous sessions. Priced at $1.00 per million input tokens and $3.00 per million output tokens, MiMo V2.5 Pro targets developers building autonomous agents, code-generation pipelines, and complex tool-use workflows where sustained coherence over long contexts is critical.

MiMo V2 Omni is Xiaomi's omni-modal foundation model that natively processes text, image, video, and audio within a unified architecture, combining multimodal perception with agentic capabilities like visual grounding, multi-step planning, and tool use. It supports over 10 hours of continuous audio understanding and a 256K context window. It outperformed Gemini 3 Pro and GPT-5.2 on several benchmarks.

MiMo V2 Pro is Xiaomi's flagship text-only reasoning model built for the 'agent era,' featuring over 1T total parameters (42B active) with a 1M-token context window, deeply optimized for agentic workflows like coding, tool calling, and task orchestration. Previously tested anonymously as 'Hunter Alpha' on OpenRouter where it topped daily API call charts, it ranks 8th globally and 2nd among Chinese LLMs on the Artificial Analysis Intelligence Index. Its agent performance approaches Claude Opus 4.6 at roughly one-fifth the cost.

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.