DeepSeek: DeepSeek V3.2
deepseek/deepseek-v3.2
Try DeepSeek V3.2 for free in your browser, and add it to your app for free with Puter.js AI API.
Try it free Add to your app// npm install @heyputer/puter.js
import { puter } from '@heyputer/puter.js';
puter.ai.chat("Explain quantum computing in simple terms", {
model: "deepseek/deepseek-v3.2"
}).then(response => {
document.body.innerHTML = response.message.content;
});
<html>
<body>
<script src="https://js.puter.com/v2/"></script>
<script>
puter.ai.chat("Explain quantum computing in simple terms", {
model: "deepseek/deepseek-v3.2"
}).then(response => {
document.body.innerHTML = response.message.content;
});
</script>
</body>
</html>
Model Card
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.
Context Window 164K
tokens
Max Output 147K
tokens
Input Cost $0.26
per million tokens
Output Cost $0.8
per million tokens
Input text
modalities
Tool Use Yes
Release Date Dec 1, 2025
Try DeepSeek V3.2 for free
Try DeepSeek V3.2 instantly in your browser.
This playground uses the Puter.js AI API — no API keys or setup required.
Benchmarks
How DeepSeek V3.2 performs on standard evaluations.
| Benchmark | Score |
|---|---|
| GPQA Diamond Graduate-level science Q&A | 75.1% |
| Humanity's Last Exam Cross-domain reasoning | 11.2% |
| LiveCodeBench Recent coding problems | 59.3% |
| AIME 2025 Advanced math exam | 59.0% |
| IFBench Instruction following | 49.0% |
| LCR Long-context reasoning | 45.7% |
| Terminal-Bench Hard Agentic terminal tasks | 32.6% |
| τ²-Bench Tool use / agents | 78.9% |
Scores sourced from Artificial Analysis.
Find other DeepSeek models →
DeepSeek V4.1 Flash
DeepSeek V4.1 Flash is DeepSeek's fast, low-cost chat model, a 552B-parameter Mixture-of-Experts system using a new causal encoder-decoder design that activates 8B parameters while reading a prompt and 16B while generating a response. It succeeds DeepSeek V4 Flash, keeping a 1,000,000-token context window and adding a continuously adjustable reasoning effort setting (1-100). DeepSeek reports it beating the larger V4 Pro on performance, cost, speed, and total runtime. In DeepSeek's own tests it scored 90.6 on Terminal-Bench 2.1, ahead of Claude Opus 5 (89.1) and GPT-5.6 Sol (88.8), 88.1 on CyberGym versus 84.5 for GPT-5.6 Sol and GLM 5.3, and 74.2 on DeepSWE v1.1 versus 74.0 for Claude Opus 5. It supports tool calling, and DeepSeek says its KV cache footprint is about a quarter of V4 Flash's for cheaper long-context serving. At $0.11/M input and $0.45/M output tokens, it fits high-throughput coding agents, tool-calling pipelines, and long-context workloads.
ChatDeepSeek V4 Flash Vision Experimental
DeepSeek V4 Flash Vision Experimental is an experimental multimodal version of DeepSeek's V4 Flash model, released by DeepSeek in August 2026. It is a sparse Mixture-of-Experts model with 284B total parameters and 13B activated per token, and it accepts images (JPEG, PNG, GIF, WebP) alongside text through the same API used for V4 Flash. DeepSeek reports that it matches V4 Flash on text tasks including agentic reasoning and world knowledge, while adding image understanding for document and chart parsing, visual question answering, and agent workflows that interleave text and images. On multimodal agent benchmarks, DeepSeek's own numbers show gains over the text-only V4 Flash, with Terminal Bench 2.1 rising from 82.7 to 83.9, ApexBench from 26.2 to 36.5, and Agents' Last Exam from 25.2 to 27.3, approaching Anthropic's Opus 4.8 (85.0 on Terminal Bench 2.1). On NL2Repo it trails Opus 4.8, 57.7 versus 69.7.
ChatDeepSeek V4 Pro 0813
DeepSeek V4 Pro 0813 is the production release of DeepSeek's V4 Pro flagship model, a 1.6-trillion-parameter Mixture-of-Experts model with 49B parameters active per token, marking the end of a four-month preview. It keeps the preview's 1,048,576-token context window, 384,000-token max output, and three reasoning modes (non-thinking, high, and max effort). DeepSeek reports large gains over the preview build: Terminal Bench 2.1 rose from 72.1 to 87.9 and DeepSWE from 12.8 to 62.7. At max reasoning effort it scores 80.6% on SWE-bench Verified, matching Gemini 3.1 Pro, along with 90.1% on GPQA Diamond, 87.5% on MMLU-Pro, and a Codeforces rating of 3,206, ahead of GPT-5.4's 3,168. Priced at $0.38 per million input tokens and $1.14 per million output tokens. Pinning the 0813 route locks an application to this exact checkpoint rather than the rolling deepseek-v4-pro alias, useful for agentic coding and long-context reasoning tasks.
Frequently Asked Questions
You can access DeepSeek V3.2 by DeepSeek through Puter.js AI API. Include the library in your web app or Node.js project and start making calls with just a few lines of JavaScript — no backend and no configuration required. You can also use it with Python or cURL via Puter's OpenAI-compatible API.
DeepSeek V3.2 is free to try with a Puter account. Every account includes a free AI allowance, and you can chat with it in the playground on this page. You can upgrade your account anytime for a larger allowance.
DeepSeek V3.2 is free to integrate using the Puter.js AI API. With the User-Pays Model, you can add AI to your app for $0, since users cover their own AI usage through their Puter account.
| Price per 1M tokens | |
|---|---|
| Input | $0.26 |
| Output | $0.8 |
DeepSeek V3.2 was created by DeepSeek and released on Dec 1, 2025.
DeepSeek V3.2 supports a context window of 164K tokens. For reference, that is roughly equivalent to 328 pages of text.
DeepSeek V3.2 can generate up to 147K tokens in a single response.
DeepSeek V3.2 accepts the following input types: text. It produces: text.
Yes, DeepSeek V3.2 supports tool use (function calling), allowing it to interact with external tools, APIs, and data sources as part of its response flow.
DeepSeek V3.2 scores 16.0 on the Artificial Analysis Intelligence Index, outperforming 59% of tracked models. On math, it scores 59.0 (outperforms 55% of models).
Yes — the DeepSeek V3.2 API works with any JavaScript framework, Node.js, or plain HTML through Puter.js. Just include the library and start building. See the documentation for more details.
Add DeepSeek V3.2 to your app for free
Developers can integrate DeepSeek V3.2 for free using the Puter.js AI API.
With the User-Pays Model, each user covers their own AI usage instead of the developer.