Mistral Large 4
- Input price
- $0.70
- Per 1 million tokens (USD)
- Output price
- $2.10
- Per 1 million tokens (USD)
- Effort levels
- 3total
- Default not stated
- Benchmark results
- 14total
- Benchmark types: 14
Model information
- API model ID
- mistral-large-4
- release date
- 2026-10-06
- effort
- Low / Med / High
- Effort source
- Based on "Mistral Large 4 reasoning effort" in Mistral reasoning documentation (docs.mistral.ai/capabilities/reasoning, retrieved 2026-11-13): mistral-large-4 / mistral-large-4-0 support low, medium, high (none, minimal, xhigh, max are not accepted). high is recommended, but the API default is not stated. These levels are for the GA version; the Preview version had none / high (migration note: none was renamed to low, high is unchanged).
- Claude Code
- unavailable
- Codex
- unavailable
- Availability notes
- Official statements about Claude Code / Codex support have not been confirmed. The official model list for Claude Code / Codex (checked 2026-11-12) contains only Anthropic / OpenAI models, respectively, so both are false.
- Availability source
- https://mistral.ai/news/mistral-large-4/
Price (USD per 1 million tokens)
Prices do not vary by effort; token consumption does.
| input | $0.70 |
|---|---|
| output | $2.10 |
| cache read | $0.07 |
| cache write | — |
Public preview at announcement. effort settings not stated
Input $0.7, cached input $0.07, output $2.1 (per 1M tokens)Benchmark results
| Benchmark | effort | Score | Source cost (metric) | measured/reported by | source | Conditions and extraction method |
|---|---|---|---|---|---|---|
| Terminal-Bench 4.0not stated | not stated | 28.3% | — | Mistral AI | Introducing Mistral Large 4official · Retrieved 2026-11-12 | Vendor-reported value. Self-reported by Mistral (as of the 2026-10-06 preview version; a separate article says the later open weights version improved with additional RL)transcribed from textSource excerptTerminal-Bench 4: 28.3% |
| DeepSWE v1.1not stated | not stated | 61.7% | — | Mistral AI | Introducing Mistral Large 4official · Retrieved 2026-11-12 | Vendor-reported value. Self-reported by Mistral (as of the 2026-10-06 preview version; a separate article says the later open weights version improved with additional RL)transcribed from textSource excerptDeepSWE v1.1: 61.7% |
| AA-Briefcase (Elo, version not stated)not stated | not stated | 1,393 | — | Mistral AI | Introducing Mistral Large 4official · Retrieved 2026-11-12 | Vendor-reported value. Self-reported by Mistral (as of the 2026-10-06 preview version; a separate article says the later open weights version improved with additional RL). AA-Briefcase version not statedtranscribed from textSource excerptAA-Briefcase: 1,393 Elo |
| AutomationBenchnot stated | not stated | 59.9% | — | Mistral AI | Introducing Mistral Large 4official · Retrieved 2026-11-12 | Vendor-reported value. Self-reported by Mistral (as of the 2026-10-06 preview version; a separate article says the later open weights version improved with additional RL)transcribed from textSource excerptAutomationBench: 59.9% |
| SWE-Atlas-QnAnot stated | not stated | 59.4% | — | Mistral AI | Introducing Mistral Large 4official · Retrieved 2026-11-12 | Vendor-reported value. Self-reported by Mistral (as of the 2026-10-06 preview version; a separate article says the later open weights version improved with additional RL)transcribed from textSource excerptSWE-Atlas-QnA: 59.4% |
| Cybenchnot stated | not stated | 93.0% | — | Mistral AI | Introducing Mistral Large 4official · Retrieved 2026-11-12 | Vendor-reported value. Self-reported by Mistral (as of the 2026-10-06 preview version; a separate article says the later open weights version improved with additional RL)transcribed from textSource excerptCybench: 93% |
| GPQA DiamondHigh | High | 88.1% | — | Epoch AI | Epoch AI, Capabilities & benchmarking (benchmark_data.zip, gpqa_diamond.csv) CC BY 4.0third-party · Retrieved 2026-11-12 | Run by Epoch AI (third-party measurement). Source label "Mistral Large 4 Preview (high)". effort is indicated by the suffix _high in Epoch's Model version. Data retrieved 2026-11-12. Epoch's label is Mistral Large 4 Preview (2026-10-06 version; Mistral Large 4 from 2026-10-29 has a different Model version, mistral-large-4-0-oss, and is not included here).transcribed from tableSource excerptgpqa_diamond.csv: Model version=mistral-large-4-0_high, Best score (across scorers)=0.8806818181818182, stderr=0.019328132475292338, Started at=2026-10-12T21:38:22.000ZLocation gpqa_diamond.csv, data row 15 (file lines 27-38; multi-line record), id=Ekx5phukyiJhAcQodMfS2p, Model version=mistral-large-4-0_high |
| FrontierMath Tiers 1-3(v2)High | High | 58.6% | — | Epoch AI | Epoch AI, Capabilities & benchmarking (benchmark_data.zip, frontiermath_tiers_1_3_v2.csv) CC BY 4.0third-party · Retrieved 2026-11-12 | Run by Epoch AI (third-party measurement). Source label "Mistral Large 4 Preview (high)". effort is indicated by the suffix _high in Epoch's Model version. Data retrieved 2026-11-12. Epoch's label is Mistral Large 4 Preview (2026-10-06 version; Mistral Large 4 from 2026-10-29 has a different Model version, mistral-large-4-0-oss, and is not included here).transcribed from tableSource excerptfrontiermath_tiers_1_3_v2.csv: Model version=mistral-large-4-0_high, Best score (across scorers)=0.5859649122807018, stderr=0.02922773823944577, Started at=2026-10-12T21:38:35.000ZLocation frontiermath_tiers_1_3_v2.csv, data row 8 (file lines 9-20; multi-line record), id=CAstp8hQG6Aehvtw85bMDX, Model version=mistral-large-4-0_high |
| FrontierMath Tier 4(v2)High | High | 26.8% | — | Epoch AI | Epoch AI, Capabilities & benchmarking (benchmark_data.zip, frontiermath_tier_4_v2.csv) CC BY 4.0third-party · Retrieved 2026-11-12 | Run by Epoch AI (third-party measurement). Source label "Mistral Large 4 Preview (high)". effort is indicated by the suffix _high in Epoch's Model version. Data retrieved 2026-11-12. Epoch's label is Mistral Large 4 Preview (2026-10-06 version; Mistral Large 4 from 2026-10-29 has a different Model version, mistral-large-4-0-oss, and is not included here).transcribed from tableSource excerptfrontiermath_tier_4_v2.csv: Model version=mistral-large-4-0_high, Best score (across scorers)=0.2682926829268293, stderr=0.07005564203095158, Started at=2026-10-12T21:38:35.000ZLocation frontiermath_tier_4_v2.csv, data row 8 (file lines 9-20; multi-line record), id=TzoNRjKto6wEuZ5PmQPAw8, Model version=mistral-large-4-0_high |
| OTIS Mock AIME 2024-2025High | High | 98.3% | — | Epoch AI | Epoch AI, Capabilities & benchmarking (benchmark_data.zip, otis_mock_aime_2024_2025.csv) CC BY 4.0third-party · Retrieved 2026-11-12 | Run by Epoch AI (third-party measurement). Source label "Mistral Large 4 Preview (high)". effort is indicated by the suffix _high in Epoch's Model version. Data retrieved 2026-11-12. Epoch's label is Mistral Large 4 Preview (2026-10-06 version; Mistral Large 4 from 2026-10-29 has a different Model version, mistral-large-4-0-oss, and is not included here).transcribed from tableSource excerptotis_mock_aime_2024_2025.csv: Model version=mistral-large-4-0_high, Best score (across scorers)=0.9833333333333333, stderr=0.006405884289881728, Started at=2026-10-12T21:38:35.000ZLocation otis_mock_aime_2024_2025.csv, data row 15 (file lines 27-38; multi-line record), id=Pby5WuP8Muu6Jzs2v5gB7F, Model version=mistral-large-4-0_high |
| SimpleQA VerifiedHigh | High | 20.0% | — | Epoch AI | Epoch AI, Capabilities & benchmarking (benchmark_data.zip, simpleqa_verified.csv) CC BY 4.0third-party · Retrieved 2026-11-12 | Run by Epoch AI (third-party measurement). Source label "Mistral Large 4 Preview (high)". effort is indicated by the suffix _high in Epoch's Model version. Data retrieved 2026-11-12. Epoch's label is Mistral Large 4 Preview (2026-10-06 version; Mistral Large 4 from 2026-10-29 has a different Model version, mistral-large-4-0-oss, and is not included here).transcribed from tableSource excerptsimpleqa_verified.csv: Model version=mistral-large-4-0_high, Best score (across scorers)=0.2, stderr=0.012655439943366657, Started at=2026-10-09T14:12:09.000ZLocation simpleqa_verified.csv, data row 17 (file lines 30-41; multi-line record), id=hvNNKdnzLvr9X8zVxWKcfk, Model version=mistral-large-4-0_high |
| Chess Puzzles(Epoch AI)High | High | 33.0% | — | Epoch AI | Epoch AI, Capabilities & benchmarking (benchmark_data.zip, chess_puzzles.csv) CC BY 4.0third-party · Retrieved 2026-11-12 | Run by Epoch AI (third-party measurement). Source label "Mistral Large 4 Preview (high)". effort is indicated by the suffix _high in Epoch's Model version. Data retrieved 2026-11-12. Epoch's label is Mistral Large 4 Preview (2026-10-06 version; Mistral Large 4 from 2026-10-29 has a different Model version, mistral-large-4-0-oss, and is not included here).transcribed from tableSource excerptchess_puzzles.csv: Model version=mistral-large-4-0_high, Best score (across scorers)=0.33, stderr=0.04725815626252608, Started at=2026-10-15T22:52:59.000ZLocation chess_puzzles.csv, data row 12 (file lines 24-35; multi-line record), id=eNdDCt2u6A8mPisK56SApU, Model version=mistral-large-4-0_high |
| MirrorCodeHigh | High | 27.8% | — | Epoch AI | Epoch AI, Capabilities & benchmarking (benchmark_data.zip, mirrorcode.csv) CC BY 4.0third-party · Retrieved 2026-11-12 | Run by Epoch AI (third-party measurement). Source label "Mistral Large 4 Preview (high)". effort is indicated by the suffix _high in Epoch's Model version. Data retrieved 2026-11-12. Epoch's label is Mistral Large 4 Preview (2026-10-06 version; Mistral Large 4 from 2026-10-29 has a different Model version, mistral-large-4-0-oss, and is not included here).transcribed from tableSource excerptmirrorcode.csv: Model version=mistral-large-4-0_high, Best score (across scorers)=0.2777777777777778, stderr=0.0885907699198465, Started at=2026-10-21T16:48:20.000ZLocation mirrorcode.csv, data row 20 (file lines 25-36; multi-line record), id=5kYFmmvF8XSX6hFMwhSxB4, Model version=mistral-large-4-0_high |
| Epoch Capabilities Index(ECI)not stated | not stated | 152.69 | — | Epoch AI | Epoch AI, epoch_capabilities_index/eci_scores.csv (CC BY 4.0)third-party · Retrieved 2026-11-12 | A per-model metric that combines results across multiple benchmarks and multiple efforts (not per effort). Calculated by Epoch AI.transcribed from tableSource excerpteci_scores.csv: Model=Mistral Large 4 Preview, eci=152.69, eci_ci_low=150.85, eci_ci_high=154.7Location epoch_capabilities_index/eci_scores.csv, data row 57 (file line 58), Model=Mistral Large 4 Preview |
Related rankings
- Overall ranking (all models, records)
- Terminal-Bench 4.0 ranking by metric (all models, records)
- DeepSWE v1.1 ranking by metric (all models, records)
- AA-Briefcase (Elo, version not stated) ranking by metric (all models, records)
- AutomationBench ranking by metric (all models, records)
- SWE-Atlas-QnA ranking by metric (all models, records)
- Cybench ranking by metric (all models, records)
- GPQA Diamond ranking by metric (all models, records)
- FrontierMath Tiers 1-3(v2) ranking by metric (all models, records)
- FrontierMath Tier 4(v2) ranking by metric (all models, records)
- OTIS Mock AIME 2024-2025 ranking by metric (all models, records)
- SimpleQA Verified ranking by metric (all models, records)
- Chess Puzzles(Epoch AI) ranking by metric (all models, records)
- MirrorCode ranking by metric (all models, records)
- Epoch Capabilities Index(ECI) ranking by metric (all models, records)
- Recommendations by role