Strengths
- Permissive Apache 2.0 license and only 6.5 billion active parameters per answer, so it is cheap and fast to run.
- Everyday tasks, reasoning and coding in one model, and thinking can be switched on per request.
- Low API price: $0.15 per million input tokens and $0.60 per million output tokens.
- Mistral says it beats gpt-oss-120b on the LiveCodeBench coding test while writing 20% shorter answers.
- A 256K token context window, and it understands both text and images.
Weaknesses
- Despite the “Small” name, the full model has 119 billion parameters. Running it yourself takes a server with several GPUs, a laptop will not do.
- The model page has no full table of test results and most comparisons appear only as charts.
Best for
- cheap high-volume tasks through the API
- a company assistant on your own server
- sorting texts and extracting data from them
- answers based on your own documents (RAG)
Not ideal for
- the hardest reasoning and coding tasks
- laptops and consumer gaming GPUs
How to use it in Europe
You can use it through the Mistral API anywhere in the EU, and data is processed in the EU by default. The weights are free to download from Hugging Face under Apache 2.0, so you can run it on your own server and keep your data in house.
Price compared with similar models
Benchmark results
| Benchmark | Score | Source |
|---|---|---|
| GPQA Diamond | 71.2 | huggingface.co ↗ (vendor-reported), March 2026 |
| AA LCR | 0.72 | huggingface.co ↗ (vendor-reported), March 2026 |
Benchmarks cover only a narrow slice of ability and vendors often report them at their best settings. Treat them as a rough guide.
More models from this vendor
- Mistral Large 4: Mistral's largest model, still in preview. Built for demanding tasks, long documents and security analysis with data kept in the EU.
- Mistral Medium 3.5: Mistral's mid-size open-weight model and the default in the Vibe app. Good for everyday work and coding with data kept in the EU.
- Ministral 3 14B: Mistral's small open model under Apache 2.0 that runs on a single GPU, even without an internet connection.
- Codestral 25.08: Mistral's model for code completion right in your editor. Fast and cheap, but dated for autonomous coding agents.
- Mistral Large 3: Mistral's previous large model with open weights under Apache 2.0. Cheap in the API and suited to your own powerful server.
Similar models from other vendors
- Claude Haiku 5.5 (Anthropic): Claude's cheapest and fastest model, built for classification, data extraction and large volumes of simple tasks.
- DeepSeek V4.1 Flash (DeepSeek): DeepSeek's cheap, fast open-weight model with a 1 million token window. In the app and API, though, your data goes to China.
- Gemini 3.5 Flash-Lite (Google): A very fast, low-cost Google model for large volumes of simple tasks, such as pulling data out of documents.
- GPT-6.1 Terra (OpenAI): OpenAI's fast, cheaper model for working with images, operating a computer and everyday office work.
- Qwen3.8-Flash (Alibaba (Qwen)): A very cheap, fast Qwen model for high volumes of tasks. Unlike Max and Plus, it runs in Frankfurt with processing kept in the EU.
- Grok 4.3 (xAI): An older, cheaper Grok from April 2026 with a 1-million-token window. Good for lots of low-cost API calls.