Model profile
Mistral Medium 3
Evidence summary
Safety. Mistral Medium 3 has an estimated Safety rank of #306; its 90% source-sensitivity interval is #179–#343. Its behavior-only rank is #298; company governance moves the combined estimate to #306. Published Safety evidence spans 6 eval lineages and 5 of 7 components. Its strongest relative result is AA-Omniscience (hallucination_rate, #107 of 358); its weakest is Concordia — SOSBench-Chem (safety_score, #63 of 63).
Freedom. Mistral Medium 3 has an estimated Freedom rank of #6; its 90% source-sensitivity interval is #2–#355. Published Freedom evidence spans 2 eval lineages and 1 of 1 components. Its strongest relative result is UGI Leaderboard — base-model willingness (willingness_adherence_score, #2 of 156); its weakest is Concordia — Fortress-Biological (safety_score, #12 of 54).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗mistral-medium-3
- OpenRouter ↗mistralai/mistral-medium-3
- Official model documentation ↗Family-level model document · mistral · first party
- Release source ↗curated outward evidence
Safety evals
Rank and direction are specific to the Safety portfolio. Black marks the observed result; the grey dot marks the value implied by the Safety rank. Values stay on each source’s native scale.
| Eval / sub-eval | Rank | Value | Distribution | Source |
|---|---|---|---|---|
| AA-Omnisciencehallucination_rate | #107 / 358 | ↓0.609 | Source ↗official | |
| Concordia — Agentic-Misalignmentsafety_score | #32 / 54 | ↑72.33 | Source ↗official | |
| Concordia — AirBench-Deceptionsafety_score | #60 / 63 | ↑37.78 | Source ↗official | |
| Concordia — AirBench-Manipulationsafety_score | #56 / 56 | ↑42.67 | Source ↗official | |
| Concordia — AirBench-SecurityRiskssafety_score | #62 / 63 | ↑41.67 | Source ↗official | |
| Concordia — APEsafety_score | #53 / 55 | ↑0.4375 | Source ↗official | |
| Concordia — CyberSecEval2-PromptInjectionsafety_score | #51 / 63 | ↑73.31 | Source ↗official | |
| Concordia — DarkBenchsafety_score | #35 / 55 | ↑53.03 | Source ↗official | |
| Concordia — Fortress-Biologicalsafety_score | #43 / 54 | ↑29.4 | Source ↗official | |
| Concordia — Fortress-Chemicalsafety_score | #42 / 54 | ↑32.88 | Source ↗official | |
| Concordia — Fortress-Privacy/Scamssafety_score | #53 / 54 | ↑30.19 | Source ↗official | |
| Concordia — MASKsafety_score | #56 / 62 | ↑42.04 | Source ↗official | |
| Concordia — SciKnowEval-BiologicalHarmfulQAsafety_score | #58 / 63 | ↑44.11 | Source ↗official | |
| Concordia — SciKnowEval-ChemicalHarmfulQAsafety_score | #56 / 63 | ↑3.965 | Source ↗official | |
| Concordia — Shutdown-Resistancesafety_score | #36 / 53 | ↑98 | Source ↗official | |
| Concordia — SOSBench-Biosafety_score | #59 / 63 | ↑11.8 | Source ↗official | |
| Concordia — SOSBench-Chemsafety_score | #63 / 63 | ↑25.6 | Source ↗official | |
| Confabulationsconfabulation_rate | #41 / 52 | ↓38.12 | Source ↗official | |
| Emergent Collusionhigh_illegality_game_rate | #8 / 13 | ↓0.47 | Source ↗official | |
| Humanity's Last Exam RMS calibration error (Scale Labs)calibrationError | #26 / 42 | ↓77 | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #87 / 131 | ↑6.68 | Source ↗official | |
| SpeciEvalland_animal_4ns | #123 / 131 | ↓5.03 | Source ↗official | |
| SpeciEvalsea_animal_4ns | #128 / 131 | ↓5.53 | Source ↗official | |
| SpeciEvalspeciesism | #113 / 131 | ↓2.62 | Source ↗official |
Freedom evals
Rank and direction are specific to the Freedom portfolio. Black marks the observed result; the grey dot marks the value implied by the Freedom rank. Values stay on each source’s native scale.
Values evaluations
Descriptive values and political-framing results are separate from safety/ethics ranks. Each strip shows the evaluation’s observed model range; its endpoint labels state what lower and higher values mean.