Mistral AIMistral Small 3.2
Best configuration ranks #510 of 610 on the BenchLeader Index at 40.3 ±1.5. Last measured 20 Jun 2025stale: no new result in six months. Released 20 Jun 2025.
- Blended price
- $0.090/M
- $0.060 in · $0.180 out
- Output speed
- 153 tok/s
- First answer
- 0.73 s
- first token 0.60 s
- Context
- 131k
- Overall index40
- Reasoning38
- Agents & tools40
- Maths37
- Knowledge40
- Instruction following37
- Multimodal32
- Long context35
- Composite38
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 | 49.0%#203 | Epoch AI Benchmarking Hub |
| GPQA Diamond (AA)not in index | 50.5%#405 | Artificial Analysis |
| Humanity's Last Exam (AA)not in index | 4.3%#429 | Artificial Analysis |
| Kagi LLM Benchmark | 40.4%#104 | Kagi LLM Benchmark |
Coding
| Benchmark | default | Source |
|---|---|---|
| SciCode (AA)not in index | 28.6%#148 | Artificial Analysis |
Agents & tools
| Benchmark | default | Source |
|---|---|---|
| Terminal-Bench Hard | 6.8%#240 | Artificial Analysis |
| τ²-Bench Telecom (AA)not in index | 29.5%#254 | Artificial Analysis |
Maths
| Benchmark | default | Source |
|---|---|---|
| OTIS Mock AIME | 30.3%#189 | Epoch AI Benchmarking Hub |
Knowledge
| Benchmark | default | Source |
|---|---|---|
| AA-Omniscience | -50.6#317 | Artificial Analysis |
Instruction following
| Benchmark | default | Source |
|---|---|---|
| IFBench | 33.5%#312 | Artificial Analysis |
Multimodal
| Benchmark | default | Source |
|---|---|---|
| MMMU-Pro | 48.0%#214 | Artificial Analysis |
Long context
| Benchmark | default | Source |
|---|---|---|
| AA-LCR | 20.3%#360 | Artificial Analysis |
Composite
| Benchmark | default | Source |
|---|---|---|
| Epoch Capabilities Indexnot in index | 131.7#127 | Epoch AI Benchmarking Hub |
| AA Intelligence Index | 7.0#417 | Artificial Analysis |
Where to run it
Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.
| Provider | Speed | First token | Input $/M | Output $/M | Context | Quantisation |
|---|---|---|---|---|---|---|
| DeepInfra | 28 tok/s | 0.56 s | $0.075 | $0.200 | 128k | fp8 |
| Parasail | 28 tok/s | 0.65 s | $0.090 | $0.300 | 131k | bf16 |
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 | $0.0001 | – | 2.7 s |
| Summarise a 30-page report | 12,000 / 600 | $0.0008 | – | 4.7 s |
| Code edit | 6,000 / 1,500 | $0.0006 | – | 10.5 s |
| Agentic coding session | 60,000 / 4,000 | $0.0043 | – | 26.9 s |
| Structured extraction | 2,000 / 200 | $0.0002 | – | 2.0 s |
See also
Data as of 9 Sept 2026. Compare with another model.