BlogLLM Benchmark Scores as Published, Side by Side
Mistral Medium 3: Benchmark Scores as Published
Every result the catalog holds for Mistral Medium 3: 22 tests in 7 areas, from 2 pages. All were printed by Mistral AI itself. Each number is as printed; nothing is ranked or rescored.
Developer Mistral AIReleased 2025-05-07Weights ClosedData version 2026103001, as of 2026-10-30
Science and Math
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| GPQA Diamond | 59.6% | Mistral AI, Announcing Magistralchecked 2026-10-15; printed on 1 more page | - |
| AIME 2025 | 21.2% | Mistral AI, Announcing Magistralchecked 2026-10-15 | Attempt budget: pass@1 |
| AIME 2024 | 26.8% | Mistral AI, Announcing Magistralchecked 2026-10-15 | Attempt budget: pass@1 |
| MATH-500 | 91% | Mistral AI, Medium is the new largechecked 2026-10-15 | - |
| AIME 2024 cons@64 (majority vote over 64 samples) | 43.3% | Mistral AI, Announcing Magistralchecked 2026-10-15 | Attempt budget: maj@64 |
| AIME 2024 maj@4 (majority vote over 4 samples) | 34.4% | Mistral AI, Announcing Magistralchecked 2026-10-15 | Attempt budget: maj@4 |
| AIME 2025 maj@4 (majority vote over 4 samples) | 25.8% | Mistral AI, Announcing Magistralchecked 2026-10-15 | Attempt budget: maj@4 |
| AIME 2025 maj@64 (majority vote over 64 samples) | 30% | Mistral AI, Announcing Magistralchecked 2026-10-15 | Attempt budget: maj@64 |
Long Context
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| RULER at 128K | 90.2% | Mistral AI, Medium is the new largechecked 2026-10-15 | Context: 128000 |
| RULER at 32K | 96% | Mistral AI, Medium is the new largechecked 2026-10-15 | Context: 32000 |
Multimodal Understanding
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| MMMU (validation) | 66.1% | Mistral AI, Medium is the new largechecked 2026-10-15 | - |
| AI2D (test) | 93.7% | Mistral AI, Medium is the new largechecked 2026-10-15 | - |
| ChartQA | 82.6% | Mistral AI, Medium is the new largechecked 2026-10-15 | - |
| DocVQA (ANLS) | 95.3 | Mistral AI, Medium is the new largechecked 2026-10-15 | - |
General Knowledge and Reasoning
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| MMLU-Pro | 77.2% | Mistral AI, Medium is the new largechecked 2026-10-15 | - |
Code Generation
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| LiveCodeBench v6 | 30.3% | Mistral AI, Medium is the new largechecked 2026-10-15 | - |
| Aider polyglot | 28.9% | Mistral AI, Announcing Magistralchecked 2026-10-15 | - |
| MultiPL-E | 81.4% | Mistral AI, Medium is the new largechecked 2026-10-15 | - |
| HumanEval | 92.1% | Mistral AI, Medium is the new largechecked 2026-10-15 | - |
| LiveCodeBench v5 | 29.1% | Mistral AI, Announcing Magistralchecked 2026-10-15 | - |
Instruction Following
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| IFEval | 89.4% | Mistral AI, Medium is the new largechecked 2026-10-15 | - |
Preference and Composite Rankings
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| Arena-Hard win rate (GPT-4-1106 judge) | 97.1% | Mistral AI, Medium is the new largechecked 2026-10-15 | - |
More from Mistral AI
Every number here is read from the same catalog file as the benchmark table, data version 2026103001. 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.