BlogLLM Benchmark Scores as Published, Side by Side
Mistral Large 4: Benchmark Scores as Published
Every result the catalog holds for Mistral Large 4: 12 tests in 8 areas, from 5 pages. 9 printed by Mistral AI itself, the rest by the company named beside each number. Each number is as printed; nothing is ranked or rescored.
Developer Mistral AIReleased 2026-10-29Context window 1MInput Text, imageOutput TextWeights Open (Mistral Ecosystem Open-Weight License)Size 1T (52B active)Runs on 8 × B300 (about 662 GB at 4-bit, estimate)Data version 2026111201, as of 2026-11-12
Knowledge Work
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| FinWorkBench (Finch) | 73.3% | Mistral AI, Mistral Large 4 (Le Chonk): Built for Coders, Trusted by Defenderschecked 2026-11-05 | - |
Agentic Coding
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| Terminal-Bench 4.0 | 41% | Mistral AI, Mistral Large 4 (Le Chonk): Built for Coders, Trusted by Defenderschecked 2026-11-05 | - |
| Vibe Code Bench | 81.56% | Vals AI, Vibe Code Bench v1.1 leaderboardchecked 2026-11-05 | Harness: OpenHands |
| FrontierCode 2.0 | 19.7% | Mistral AI, Mistral Large 4 (Le Chonk): Built for Coders, Trusted by Defenderschecked 2026-11-05 | - |
| Artificial Analysis Coding Agent Index (version not stated) | 48 | Mistral AI, Mistral Large 4 (Le Chonk): Built for Coders, Trusted by Defenderschecked 2026-11-05 | - |
Science and Math
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| GPQA Diamond | 90.2% | Epoch AI, Epoch AI Benchmarking Hubchecked 2026-11-05 | Effort: high |
Cybersecurity
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| Artificial Analysis Cyber Index (successes) | 51 | Mistral AI, Mistral Large 4 (Le Chonk): Built for Coders, Trusted by Defenderschecked 2026-11-05 | - |
General Knowledge and Reasoning
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| SimpleQA Verified | 22.4% | Epoch AI, Epoch AI Benchmarking Hubchecked 2026-11-05 | Effort: high |
Web Search and Research
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| BrowseComp | 84.5% | Mistral AI, Mistral Large 4 (Le Chonk): Built for Coders, Trusted by Defenderschecked 2026-11-05 | - |
Safety and Alignment
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| B3 AI Security Benchmark (attack resistance) | 97.1% | Mistral AI, Mistral Large 4 (Le Chonk): Built for Coders, Trusted by Defenderschecked 2026-11-05 | - |
| Cyber refusal rate (average of JailbreakBench, StrongREJECT, AgentHarm) | 96.6% | Mistral AI, Mistral Large 4 (Le Chonk): Built for Coders, Trusted by Defenderschecked 2026-11-05 | - |
Preference and Composite Rankings
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| Artificial Analysis Intelligence Index v5.0 | 26 | Mistral AI, Mistral Large 4 model card (Hugging Face)checked 2026-11-05 | - |
Price
API list prices in US dollars per one million tokens, as printed on the developer's pricing page on the day checked. Cost per task is measured by Vals AI.
| Price | Value | Printed by | Setup as printed |
|---|---|---|---|
| Input | $0.7 | Mistral AI, Mistral API pricingchecked 2026-11-05 | - |
| Cached input | $0.07 | Mistral AI, Mistral API pricingchecked 2026-11-05 | - |
| Output | $2.1 | Mistral AI, Mistral API pricingchecked 2026-11-05 | - |
| Cost per taskVals AI, USD, lower is better | $11.72 | Vals AI, Vibe Code Bench v1.1 leaderboardchecked 2026-11-05 | Harness: OpenHands |
Side by side
More from Mistral AI
Every number here is read from the same catalog file as the benchmark table, data version 2026111201. Where one page printed the model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.