Mistral AIAuto-detectedMagistral Small 1.2
Magistral Small 1.2 is a Mistral AI proprietary model, released 18 Sept 2025. Its best configuration ranks #505 of 372 on the BenchLeader Index at 41.2 ±5.6, in the lower half. It scores highest in coding (47) and lowest in knowledge (28). At $0.750 per million tokens blended it is mid-priced. Last measured 10 Sept 2026.
- Blended price
- $0.750/M
- $0.500 in · $1.50 out
- Output speed
- –
- First answer
- –
- Context
- 128k
- Overall index41
- Reasoning37
- Coding47
- Agents & tools38
- Maths47
- Knowledge28
- Instruction following46
- Multimodal39
- Long context35
- Composite40
Versions
Mistral AI has shipped 5 models under this name. Each is ranked on its own results; a newer version often has fewer results so far, which holds its index nearer the average until more arrive.
| Model | Released | Index | Rank |
|---|---|---|---|
| Magistral Small 1.2this page | 18 Sept 2025 | 41.2 | #505 |
| Magistral Small 1.0 | 10 Jun 2025 | – | – |
| Magistral Small | 17 Mar 2025 | 37.9 | #570 |
| Magistral Small 1.0 | – | – | – |
| Magistral Small 1.2 | – | – | – |
Benchmark results
One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.
Reasoning
| Benchmark | default | Source |
|---|---|---|
| GPQA Diamond | 47.6%#212 | Epoch AI Benchmarking Hub |
| GPQA Diamond (AA)not in index | 66.3%#324 | Artificial Analysis |
| Humanity's Last Exam (AA)not in index | 6.4%#323 | Artificial Analysis |
| GPQA Diamond (Vals)not in index | 58.3%#108 | Vals AI |
Coding
| Benchmark | default | Source |
|---|---|---|
| SciCode | 35.2%#137 | SciCode |
| LiveCodeBench | 72.1%#82 | Vals AI |
Agents & tools
| Benchmark | default | Source |
|---|---|---|
| Terminal-Bench Hard | 4.5%#275 | Artificial Analysis |
| τ²-Bench Telecom (AA)not in index | 27.8%#264 | Artificial Analysis |
Maths
| Benchmark | default | Source |
|---|---|---|
| OTIS Mock AIME | 28.1%#197 | Epoch AI Benchmarking Hub |
| AIME (Vals) | 80.7%#47 | Vals AI |
| MGSM | 86.3%#59 | Vals AI |
Knowledge
| Benchmark | default | Source |
|---|---|---|
| AA-Omniscience | -65.1#431 | Artificial Analysis |
| MMLU-Pro | 62.1%#128 | Vals AI |
| LegalBench | 40.0%#136 | Vals AI |
| CorpFin | 44.0%#113 | Vals AI |
| TaxEval | 60.3%#124 | Vals AI |
| MedQA | 82.4%#63 | Vals AI |
Instruction following
| Benchmark | default | Source |
|---|---|---|
| IFBench | 44.4%#209 | Artificial Analysis |
Multimodal
| Benchmark | default | Source |
|---|---|---|
| MMMU-Pro | 55.5%#195 | Artificial Analysis |
Long context
| Benchmark | default | Source |
|---|---|---|
| AA-LCR | 19.3%#370 | Artificial Analysis |
Composite
| Benchmark | default | Source |
|---|---|---|
| AA Intelligence Index | 8.6#364 | Artificial Analysis |
What a task costs
Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0007 | – | – |
| Summarise a 30-page report | 12,000 / 600 | $0.0069 | – | – |
| Code edit | 6,000 / 1,500 | $0.0053 | – | – |
| Agentic coding session | 60,000 / 4,000 | $0.036 | – | – |
| Structured extraction | 2,000 / 200 | $0.0013 | – | – |
See also
Data as of 10 Sept 2026. Compare with another model.