Mistral Small 3.1 (25.03)
Single-node inference model with 128k context window supporting dozens of languages and 80+ coding languages.
Mistral Multilingual small model with 128k context
Mistral Small 3.1 (25.03) is a text generation model developed by Mistral, released in March 2025. It features a 128,000-token context window, multimodal understanding, and support for dozens of spoken languages alongside more than 80 coding languages. The model is designed to run on a single node, making it practical for deployment without distributed infrastructure.
This version introduces improved text performance and expanded context handling compared to earlier Mistral Small releases. At an inference speed of approximately 150 tokens per second, it is suited for tasks that require both throughput and long-context processing, such as document analysis, multilingual applications, and code generation. Its combination of broad language coverage and single-node efficiency makes it a practical choice for developers building production applications with constrained compute budgets.
What Mistral Small 3.1 (25.03) supports
Long Context Window
Processes up to 128,000 tokens in a single request, enabling analysis of long documents, codebases, or extended conversations without truncation.
Multilingual Text
Supports dozens of spoken languages for generation and comprehension tasks, making it suitable for international and localized applications.
Code Generation
Handles code tasks across 80+ programming languages, including generation, completion, and explanation.
Multimodal Understanding
Accepts image inputs alongside text, allowing the model to reason about visual content within a single prompt.
Fast Inference
Delivers approximately 150 tokens per second, supporting latency-sensitive production workloads on a single node.
Function Calling
Supports structured tool use and function calling, enabling integration with external APIs and agentic workflows.
Ready to build with Mistral Small 3.1 (25.03)?
Get Started FreeBenchmark scores
Scores represent accuracy — the percentage of questions answered correctly on each test.
| Benchmark | What it tests | Score |
|---|---|---|
| MMLU-Pro | Expert knowledge across 14 academic disciplines | 52.9% |
| GPQA Diamond | PhD-level science questions (biology, physics, chemistry) | 38.1% |
| MATH-500 | Undergraduate and competition-level math problems | 56.3% |
| AIME 2024 | American math olympiad problems | 6.3% |
| LiveCodeBench | Real-world coding tasks from recent competitions | 14.1% |
| HLE | Questions that challenge frontier models across many domains | 4.3% |
| SciCode | Scientific research coding and numerical methods | 15.6% |
Common questions about Mistral Small 3.1 (25.03)
What is the context window size for Mistral Small 3.1 (25.03)?
The model supports a context window of 128,000 tokens, allowing it to process long documents or extended conversations in a single request.
Does Mistral Small 3.1 (25.03) support image inputs?
Yes. This version includes multimodal understanding, meaning it can accept and reason about image inputs in addition to text.
How many coding languages does this model support?
The model supports over 80 coding languages, making it broadly applicable for code generation, completion, and explanation tasks.
What is the knowledge cutoff date for this model?
A specific training data cutoff date is not listed in the available metadata for this model version.
Can this model run on a single machine?
Yes. Mistral Small 3.1 (25.03) is designed for single-node inference, meaning it does not require distributed compute infrastructure to run.
Parameters & options
Explore similar models
Mistral Ministral 3 3B
Designed for edge deployment, it delivers high performance across diverse hardware, including local setups.
Mistral Ministral 3 8B
Built for edge deployment, it delivers high performance across diverse hardware, including local setups.
Mistral Ministral 3 14B
Optimized for local deployment, it delivers high performance across diverse hardware, including local setups.
Google Gemini 3.1 Pro
Google's frontier reasoning model delivering enhanced software engineering, agentic reliability, and multimodal intelligence across a 1M-token context window.
OpenAI GPT‑5.2 Pro
The most capable model series yet for professional knowledge work.
Nvidia Nemotron 3 Nano 30B
NVIDIA's compact open-weight hybrid MoE model with 30B total / 3.5B active parameters, delivering strong reasoning and coding performance up to 1M context.
Start building with Mistral Small 3.1 (25.03)
No API keys required. Create AI-powered workflows with Mistral Small 3.1 (25.03) in minutes — free.