Mistral AIMistral Small 4
Mistral Small 4 is a Mistral AI open-weights reasoning model, released 16 Mar 2026. Its best configuration (thinking reasoning effort) ranks #511 of 785 on the BenchLeader Index at 45.6 ±4.0, in the lower half. The ± is the point: 299 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in instruction following (50) and lowest in agents & tools (38). At $0.262 per million tokens blended it is cheaper than most ranked models. Output speed of 183 tokens per second puts it in the fastest quarter, with a first answer in 11.8 s. It has been measured at 4 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 16 Mar 2026.stale: no new result in six months
- Blended price
- $0.262/M
- $0.150 in · $0.600 out
- Output speed
- 183 tok/s
- measured by Artificial Analysis
- First answer
- 12 s
- first token 0.52 s
- Context
- 262k
- Full answer
- 15 s
- median, reasoning included
- Cost per run
- $0.015
- one full Intelligence Index run
- Released
- 16 Mar 2026
- Overall index46
- Reasoning41
- Agents & tools38
- Knowledge48
- Instruction following50
- Multimodal39
- Long context49
- Composite42
Versions
Mistral AI has shipped 6 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 |
|---|---|---|---|
| Mistral Small 4thinkingthis page | 16 Mar 2026 | 45.6 | #511 |
| Mistral Small 3.2 | 20 Jun 2025 | 41.6 | #622 |
| Mistral Small 3.1 | 17 Mar 2025 | 36.3 | #749 |
| Mistral Small 3 | 25 Jan 2025 | 39.3 | #688 |
| Mistral Small v24.09 | 17 Sept 2024 | – | – |
| Mistral Small 2402 | 26 Feb 2024 | 32.0 | #781 |
Reasoning-effort configurations
The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Not stated” collects results from publishers that did not say which setting they used; for a reasoning model that is usually its thinking mode, but we do not assume it. Pick a setting here or at the top of the page to see its price, speed and category scores.
| Effort | Index | Rank | Speed | First answer | Chat reply cost | Categories |
|---|---|---|---|---|---|---|
| no reasoning | 41.7 | #621 | 167 tok/s | 0.94 s | $0.0002 | Agents & tools 43 · Composite 40 · Instruction following 36 · Knowledge 40 · Long context 38 · Multimodal 29 · Reasoning 41 |
| high | – | – | 104 tok/s | 0.52 s | $0.0002 | |
| thinkingbest | 45.6 | #511 | 183 tok/s | 12 s | $0.0002 | Agents & tools 38 · Composite 42 · Instruction following 50 · Knowledge 48 · Long context 49 · Multimodal 39 · Reasoning 41 |
| not stated | – | – | 104 tok/s | 0.52 s | $0.0002 | Reasoning 39 |
How its index has moved
9 Sept 2026 to 21 Oct 2026- thinking
- no reasoning
The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.
Benchmark results
See all LLM benchmarks. 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 | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| GPQA Diamond (AA)not in index | 57.1%#396 | 76.9%#227 | – | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 3.8%#564 | 9.9%#326 | – | Artificial Analysis | |
| Kagi LLM Benchmarkeffort not stated | – | – | 40.5%#103 | Kagi LLM Benchmark | |
| CritPt | 0.3%#271 | 0.3%#271 | – | Artificial Analysis | |
| LMCAnot in indexeffort not stated | – | – | 20.6%#158 | Epoch AI Benchmarking Hub | |
| DTBenchnot in indexeffort not stated | – | – | 70.9%#156 | Epoch AI Benchmarking Hub |
Coding
| Benchmark | high | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| SciCode (AA)not in index | – | 38.8%#184 | – | Artificial Analysis | |
| Vibe Code Bench v1.1not in index | 0.0%#111 | – | – | Vals AI | |
| ALE-Benchnot in indexeffort not stated | – | – | 497.6#104 | Epoch AI Benchmarking Hub | |
| Terminal-Bench 4.0 (AA)not in index | – | 0.0%#152 | – | Artificial Analysis | |
| Terminal-Bench 2.1 (AA)not in index | – | 21.0%#173 | – | Artificial Analysis |
Agents & tools
| Benchmark | no reasoning | high | thinking | Source | Trend |
|---|---|---|---|---|---|
| Terminal-Bench Hard | 10.6%#221 | – | 17.4%#178 | Artificial Analysis | |
| τ²-Bench Telecom (AA)not in index | 18.4%#347 | – | 41.2%#220 | Artificial Analysis | |
| GDPval-AA v2.1not in index | – | – | 0.0%#124 | Artificial Analysis | |
| τ³-Banking (AA) | – | – | 5.0%#185 | Artificial Analysis | |
| Analyst Agent (AA)not in index | – | – | 1.3%#49 | Artificial Analysis | |
| Terminal-Bench 2.0 (Vals)not in index | – | 16.9%#59 | – | Vals AI | |
| AutomationBenchnot in index | – | – | 1.0%#197 | Artificial Analysis | |
| GDP.pdf (AA)not in index | – | – | 0.0%#220 | Artificial Analysis |
Knowledge
| Benchmark | no reasoning | thinking | Source | Trend |
|---|---|---|---|---|
| AA-Omniscience | -48.5#376 | -30.4#283 | Artificial Analysis | |
| AA-Omniscience: accuracynot in index | 16.6%#387 | 21.7%#282 | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in index | 22.0%#268 | 33.5%#199 | Artificial Analysis |
Instruction following
| Benchmark | no reasoning | thinking | Source | Trend |
|---|---|---|---|---|
| IFBench | 32.8%#336 | 48.2%#185 | Artificial Analysis |
Multimodal
| Benchmark | no reasoning | thinking | Source | Trend |
|---|---|---|---|---|
| MMMU-Pro | 46.5%#251 | 56.8%#222 | Artificial Analysis |
Long context
| Benchmark | no reasoning | thinking | Source | Trend |
|---|---|---|---|---|
| AA-LCR | 28.3%#398 | 49.7%#325 | Artificial Analysis | |
| MLCRnot in index | – | 1.7%#162 | Artificial Analysis |
Composite
| Benchmark | no reasoning | thinking | Source | Trend |
|---|---|---|---|---|
| AA Intelligence Index v4.3.2 | 9.0#416 | 11.3#356 | 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 |
|---|---|---|---|---|---|---|
| Mistral | 111 tok/s | 0.54 s | $0.150 | $0.600 | 262k | – |
| Mistral | 91 tok/s | 0.85 s | $0.165 | $0.660 | 262k | – |
| Mistral | 86 tok/s | 0.54 s | $0.150 | $0.600 | 262k | – |
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 at $0.015 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0002 | $0.0002 | 13.5 s |
| Summarize a 30-page report | 12,000 / 600 | $0.0022 | $0.0009 | 15.1 s |
| Code edit | 6,000 / 1,500 | $0.0018 | $0.0012 | 20.0 s |
| Agentic coding session | 60,000 / 4,000 | $0.011 | $0.0053 | 33.7 s |
| Structured extraction | 2,000 / 200 | $0.0004 | $0.0002 | 12.9 s |
More from Mistral AI
See also
Questions people ask
- What are Mistral Small 4's benchmark scores?
- As of 24 Oct 2026, Mistral Small 4 (thinking effort) scores 45.6 ±4.0 on the BenchLeader Index, strongest in instruction following (50) and weakest in agents & tools (38). That rests on 8 independent benchmark results, and every score on this page links to its publisher.
- How does Mistral Small 4 rank?
- Mistral Small 4 (thinking effort) is #511 of 785 ranked configurations on 24 Oct 2026. 299 other configurations score within its margin, so the exact position matters less than the band. The leader is Claude Opus 5.5 (max effort) at 72.6.
- How much does Mistral Small 4 cost per 1M tokens?
- As of 24 Oct 2026, Mistral Small 4 costs $0.150 per 1M input tokens and $0.600 per 1M output tokens, $0.262 blended (three input tokens per output token).
- When was Mistral Small 4 released?
- Mistral AI released Mistral Small 4 on 16 Mar 2026. BenchLeader has results for it at 4 reasoning-effort settings.
Data as of 24 Oct 2026. Compare these configurations.
Cite as: BenchLeader, “Mistral Small 4: benchmarks, pricing, speed and rank”, data as of 24 Oct 2026, https://www.benchleader.com/models/mistral-small-4.