Mistral AI: Mistral Large 3
mistralai/mistral-large-2512
Try Mistral Large 3 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: "mistralai/mistral-large-2512"
}).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: "mistralai/mistral-large-2512"
}).then(response => {
document.body.innerHTML = response.message.content;
});
</script>
</body>
</html>
Model Card
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.
Context Window 262K
tokens
Max Output 262K
tokens
Input Cost $0.5
per million tokens
Output Cost $1.5
per million tokens
Input text, image
modalities
Tool Use Yes
Knowledge Cutoff Dec 2025
Release Date Dec 2, 2025
Try Mistral Large 3 for free
Try Mistral Large 3 instantly in your browser.
This playground uses the Puter.js AI API — no API keys or setup required.
Benchmarks
How Mistral Large 3 performs on standard evaluations.
| Benchmark | Score |
|---|---|
| GPQA Diamond Graduate-level science Q&A | 68.0% |
| Humanity's Last Exam Cross-domain reasoning | 4.2% |
| LiveCodeBench Recent coding problems | 46.5% |
| SciCode Scientific programming | 36.6% |
| AIME 2025 Advanced math exam | 38.0% |
| IFBench Instruction following | 36.2% |
| LCR Long-context reasoning | 36.0% |
| Terminal-Bench Hard Agentic terminal tasks | 15.9% |
| τ²-Bench Tool use / agents | 24.6% |
Scores sourced from Artificial Analysis.
Find other Mistral AI models →
Mistral Large 4
Mistral Large 4 is a multimodal sparse mixture-of-experts model from Mistral AI with 1 trillion total parameters and 49 billion active. It accepts text and image input and produces text output, and it supports tool calling and structured responses through the API. Mistral positions it for coding, cybersecurity, finance, manufacturing, and visual grounding. In Mistral's preview results it scores 62% on DeepSWE v1.1, 67% on FinWorkBench, and 73% on DIOR-RSVG, and it solves 93% of the 40 Cybench exercises. It is suited to developers building coding agents, security analysis tools, and enterprise workflows that need long context (1M tokens) and image understanding. Open weights are planned for release after the API preview.
ChatMistral Medium 3.5
Mistral Medium 3.5 is a dense 128-billion-parameter multimodal model from Mistral AI that unifies instruction-following, reasoning, and coding into a single set of weights. It features a 256k-token context window, native function calling, structured JSON output, and vision capabilities via a custom-trained encoder that handles variable image sizes. A per-request reasoning_effort parameter lets you toggle between fast responses and deeper chain-of-thought processing, making the same model suitable for quick chat replies and complex agentic workflows. On benchmarks, it scores 77.6% on SWE-Bench Verified and 91.4% on τ³-Telecom. It replaces Mistral's previous Medium 3.1, Magistral, and Devstral 2 models. Priced at $1.50 per million input tokens and $7.50 per million output tokens, it's a strong fit for developers building tool-calling agents, long-horizon coding tasks, and multi-step automation pipelines.
ChatMistral Medium 3.5
Mistral Medium 3.5 is a dense 128-billion-parameter multimodal model from Mistral AI that unifies instruction-following, reasoning, and coding into a single set of weights. This entry is Mistral's own dated direct-integration id for the same release available as mistralai/mistral-medium-3-5 through OpenRouter. It features a 256k-token context window, native function calling, structured JSON output, and vision capabilities via a custom-trained encoder that handles variable image sizes. A per-request reasoning_effort parameter lets you toggle between fast responses and deeper chain-of-thought processing, making the same model suitable for quick chat replies and complex agentic workflows. On benchmarks, it scores 77.6% on SWE-Bench Verified and 91.4% on τ³-Telecom. It replaces Mistral's previous Medium 3.1, Magistral, and Devstral 2 models. Priced at $1.50 per million input tokens and $7.50 per million output tokens, it's a strong fit for developers building tool-calling agents, long-horizon coding tasks, and multi-step automation pipelines.
Frequently Asked Questions
You can access Mistral Large 3 by Mistral AI 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.
Mistral Large 3 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.
Mistral Large 3 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 | $0.5 |
| Output | $1.5 |
Mistral Large 3 was created by Mistral AI and released on Dec 2, 2025.
Mistral Large 3 supports a context window of 262K tokens. For reference, that is roughly equivalent to 524 pages of text.
Mistral Large 3 can generate up to 262K tokens in a single response.
Mistral Large 3 has a knowledge cutoff date of Dec 2025. This means the model was trained on data available up to that date.
Mistral Large 3 accepts the following input types: text, image. It produces: text.
Yes, Mistral Large 3 supports tool use (function calling), allowing it to interact with external tools, APIs, and data sources as part of its response flow.
Mistral Large 3 scores 9.3 on the Artificial Analysis Intelligence Index, outperforming 38% of tracked models. On coding, it scores 20.1 (outperforms 20% of models). On math, it scores 38.0 (outperforms 38% of models).
Yes — the Mistral Large 3 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 Mistral Large 3 to your app for free
Developers can integrate Mistral Large 3 for free using the Puter.js AI API.
With the User-Pays Model, each user covers their own AI usage instead of the developer.