Mistral AIMistral Large 4
Mistral Large 4 is a Mistral AI open-weights reasoning model, released 6 Oct 2026. Its best configuration ranks #159 of 760 on the BenchLeader Index at 57.8 ±5.7, in the upper half. The ± is the point: 274 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 composite (73) and lowest in agents & tools (33). At $2.06 per million tokens blended it is pricier than most ranked models. Output speed of 55 tokens per second puts it slower than most, with a first token in 1.8 s. It has been measured at 2 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 8 Oct 2026.
- Blended price
- $2.06/M
- $1.36 in · $4.18 out
- Output speed
- 55 tok/s
- OpenRouter traffic, 7-day median; not yet measured by Artificial Analysis
- First answer
- 1.78 s
- Context
- 1.0M
- Full answer
- –
- Cost per run
- $1.13
- one full Intelligence Index run
- Released
- 6 Oct 2026
- Overall index58
- Reasoning60
- Coding60
- Agents & tools33
- Maths67
- Knowledge59
- Human preference61
- Multimodal60
- Long context66
- Composite73
Versions
Mistral AI has shipped 4 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 |
|---|---|---|---|
| NewMistral Large 4this page | 6 Oct 2026 | 57.8 | #159 |
| Mistral Large 3 | 1 Nov 2024 | 45.5 | #489 |
| Mistral Large 2 | 1 Nov 2024 | 40.6 | #632 |
| Mistral Large 2402 | 26 Feb 2024 | 41.0 | #622 |
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 |
|---|---|---|---|---|---|---|
| high | – | – | 55 tok/s | 1.78 s | $0.0018 | Composite 41 · Knowledge 33 |
| not statedbest | 57.8 | #159 | 55 tok/s | 1.78 s | $0.0018 | Agents & tools 33 · Coding 60 · Composite 73 · Human preference 61 · Knowledge 59 · Long context 66 · Maths 67 · Multimodal 60 · Reasoning 60 |
How its index has moved
6 Oct 2026 to 9 Oct 2026The 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
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 | high | not stated | Source | Trend |
|---|---|---|---|---|
| LMArena Hard Promptseffort not stated | – | 1453#104 | LMArena | |
| LiveBench Reasoningnot in index | 83.9%#41 | – | LiveBench | |
| Humanity's Last Exam (AA)not in indexeffort not stated | – | 35.0%#113 | Artificial Analysis | |
| CritPteffort not stated | – | 10.6%#108 | Artificial Analysis |
Coding
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| LMArena Codingeffort not stated | – | 1492#86 | LMArena | |
| LMArena WebDeveffort not stated | – | 1541#41 | LMArena | |
| LiveBench Codingnot in index | 77.2%#38 | – | LiveBench | |
| SciCode (AA)not in indexeffort not stated | – | 54.2%#61 | Artificial Analysis | |
| Terminal-Bench 4.0 (AA)not in indexeffort not stated | – | 26.8%#45 | Artificial Analysis |
Agents & tools
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| LMArena Agenteffort not stated | – | -6.6#41 | LMArena | |
| LiveBench Agentic Codingnot in index | 57.2%#21 | – | LiveBench | |
| GDPval-AA v2.1not in indexeffort not stated | – | 46.2%#64 | Artificial Analysis | |
| AutomationBenchnot in indexeffort not stated | – | 59.9%#22 | Artificial Analysis | |
| GDP.pdfnot in indexeffort not stated | – | 18.6%#31 | Artificial Analysis | |
| Harvey LABnot in indexeffort not stated | – | 3.1%#10 | Harvey | |
| AA-Briefcase v1.1not in indexeffort not stated | – | 1393#37 | Artificial Analysis |
Maths
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| LiveBench Mathematicsnot in index | 93.6%#20 | – | LiveBench | |
| LMArena Mathseffort not stated | – | 1487#30 | LMArena |
Knowledge
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| SimpleQA Verified | 20.0%#70 | – | Epoch AI Benchmarking Hub | |
| LiveBench Data Analysisnot in index | 76.5%#31 | – | LiveBench | |
| AA-Omniscienceeffort not stated | – | -5.3#157 | Artificial Analysis | |
| AA-Omniscience: accuracynot in indexeffort not stated | – | 25.8%#211 | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in indexeffort not stated | – | 58.0%#79 | Artificial Analysis |
Instruction following
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| LiveBench Languagenot in index | 49.6%#66 | – | LiveBench | |
| LiveBench Instruction Followingnot in index | 64.8%#44 | – | LiveBench | |
| LMArena Instruction Followingnot in indexeffort not stated | – | 1426#97 | LMArena |
Human preference
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| LMArena Texteffort not stated | – | 1429#109 | LMArena | |
| LMArena Creative Writingnot in indexeffort not stated | – | 1367#145 | LMArena | |
| LMArena Multi-turnnot in indexeffort not stated | – | 1425#117 | LMArena | |
| LMArena Longer Queriesnot in indexeffort not stated | – | 1435#111 | LMArena |
Multimodal
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| MMMU-Proeffort not stated | – | 76.4%#84 | Artificial Analysis |
Long context
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| AA-LCReffort not stated | – | 81.3%#52 | Artificial Analysis |
Composite
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| LiveBench | 71.8%#51 | – | LiveBench | |
| AA Intelligence Index v4.3.2effort not stated | – | 38.4#70 | Artificial Analysis |
Safety & honesty
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| FORTRESSnot in indexeffort not stated | – | 23.1%#34 | Scale AI SEAL |
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 | 79 tok/s | 0.79 s | $0.680 | $2.09 | 1.0M | – |
| Mistral | 70 tok/s | 1.60 s | $0.680 | $2.09 | 1.0M | – |
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 $0.070 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0018 | $0.0014 | 7.2 s |
| Summarise a 30-page report | 12,000 / 600 | $0.019 | $0.0072 | 12.7 s |
| Code edit | 6,000 / 1,500 | $0.014 | $0.0086 | 29.1 s |
| Agentic coding session | 60,000 / 4,000 | $0.098 | $0.040 | 1.2 min |
| Structured extraction | 2,000 / 200 | $0.0036 | $0.0016 | 5.4 s |
See also
Data as of 11 Oct 2026. Compare these configurations.
Cite as: BenchLeader, “Mistral Large 4: benchmarks, pricing, speed and rank”, https://www.benchleader.com/models/mistral-large-4, data as of 11 Oct 2026.