Magistral Small
Mistral·
LLMsopen-weightEdge / Consumer
Overview
Mistral's first reasoning model, released as open weights at 24B under Apache 2.0. The larger Magistral Medium sibling announced the same day is API-only and has no published weights — it is tracked as a separate cloud entry.
Capabilities and innovations
Reasoning traces in the response128K context windowMultilingualApache 2.0 licenseFirst reasoning model from Mistral
Benchmarks
No benchmark values are recorded for this model.
- GPQA Diamond: independent evaluation pending — The values previously carried here (GPQA 72.4, HLE 19.8, MMLU 88.4, SWE-bench 48.9) were unsourced and match the API-only Magistral Medium, not the open 24B. Re-source from the Magistral release before entering a value.
- Humanity’s Last Exam (no tools): independent evaluation pending
- MMLU: independent evaluation pending
- SWE-bench Verified: independent evaluation pending
- SWE-bench Pro: not reported by the vendor
- MMLU-Pro: not reported by the vendor
- Terminal-Bench 2.x: benchmark did not exist at release
- MMMU-Pro: benchmark did not exist at release
Architecture and hardware
- Parameters
- 24B total (Dense)
- Estimated VRAM at Q4
- ~14 GB, Edge / Consumer class
- Quantization formats
- GGUF, GPTQ, AWQ
- Recommended runtime
- Ollama
- License
- Apache 2.0
Links
More from Mistral
Data curated by AI Model Timeline. See the methodology for admission criteria, benchmark eras and source priority.