Model profile
Mistral Small 3.2
Evidence summary
Mistral Small 3.2 has an estimated overall rank of #231; its 90% source-sensitivity interval is #84–#289. Its behavior-only rank is #220; company governance moves the combined estimate to #231. Published evidence spans 3 evals and 5 of 7 behavior components. Its strongest relative result is PHARE (bias_resistance_diagnostic, #4 of 66); its weakest is Alignment Leaderboard (honesty, #24 of 24).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗mistral-small-3-2
- Hugging Face ↗mistralai/Mistral-Small-3.2-24B-Instruct-2506
- Official model page ↗Exact model document · Reviewed official Hugging Face owner · official repository
- Release source ↗repository created at proxy
Published eval results
Rank is within that sub-eval. Black marks the observed result; the grey dot marks the value implied by the global rank. Values stay on each source’s native scale.
| Eval / sub-eval | Rank | Value | Distribution | Source |
|---|---|---|---|---|
| AA-Omnisciencehallucination_rate | #132 / 328 | ↓0.7681 | Source ↗official | |
| Alignment Leaderboardcorrigibility | #23 / 24 | ↑3.854 | Source ↗official | |
| Alignment Leaderboardhonesty | #24 / 24 | ↑3.055 | Source ↗official | |
| Alignment Leaderboardnon_manipulation | #23 / 24 | ↑2.569 | Source ↗official | |
| Alignment Leaderboardrobustness | #21 / 24 | ↑3.08 | Source ↗official | |
| Alignment Leaderboardsafety | #22 / 24 | ↑3.038 | Source ↗official | |
| Alignment Leaderboardscheming | #23 / 24 | ↑3.241 | Source ↗official | |
| PHAREbias_resistance_diagnostic | #4 / 66 | ↑0.739 | Source ↗official | |
| PHAREhallucination_resistance_diagnostic | #61 / 70 | ↑0.634 | Source ↗official | |
| PHAREharm_resistance_diagnostic | #55 / 70 | ↑0.8787 | Source ↗official | |
| PHAREjailbreak_resistance_diagnostic | #61 / 67 | ↑0.3289 | Source ↗official |
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.