Model profile
Mistral Large
Evidence summary
Safety. Mistral Large has an estimated Safety rank of #87; its 90% source-sensitivity interval is #36–#269. Its behavior-only rank is #75; company governance moves the combined estimate to #87. Published Safety evidence spans 6 eval lineages and 6 of 7 components. Its strongest relative result is AnimalHarmBench (score, #1 of 10); its weakest is OR-Bench (toxic_acceptance_rate, #25 of 25).
Freedom. Mistral Large has an estimated Freedom rank of #187; its 90% source-sensitivity interval is #77–#454. Published Freedom evidence spans 5 eval lineages and 1 of 1 components. Its strongest relative result is OR-Bench (toxic_acceptance_rate, #1 of 25); its weakest is BlueBench AttaQ-100 (attaq_harmlessness_reward_pct, #17 of 18).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗mistral-large
- OpenRouter ↗mistralai/mistral-large
- Model card ↗Family-level model document · Mistral AI · first party
- Release source ↗direct research preview identity date
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 |
|---|---|---|---|---|
| Adversarial Robustnessscore | #7 / 8 | ↓37 | Source ↗official | |
| AILuminate General Purpose AI Chatchild_sexual_exploitation | #17 / 32 | ↓21.5 | Source ↗official | |
| AILuminate General Purpose AI Chatdefamation | #24 / 32 | ↓26.3 | Source ↗official | |
| AILuminate General Purpose AI Chathate | #18 / 32 | ↓9.1 | Source ↗official | |
| AILuminate General Purpose AI Chatindiscriminate_weapons_cbrne | #22 / 32 | ↓31.9 | Source ↗official | |
| AILuminate General Purpose AI Chatintellectual_property | #21 / 32 | ↓18 | Source ↗official | |
| AILuminate General Purpose AI Chatnon_violent_crimes | #26 / 32 | ↓31.1 | Source ↗official | |
| AILuminate General Purpose AI Chatprivacy | #25 / 32 | ↓25.9 | Source ↗official | |
| AILuminate General Purpose AI Chatsex_related_crimes | #23 / 32 | ↓21.2 | Source ↗official | |
| AILuminate General Purpose AI Chatsexual_content | #22 / 32 | ↓34.3 | Source ↗official | |
| AILuminate General Purpose AI Chatspecialized_advice | #23 / 32 | ↓22.7 | Source ↗official | |
| AILuminate General Purpose AI Chatsuicide_self_harm | #14 / 32 | ↓12 | Source ↗official | |
| AILuminate General Purpose AI Chatviolent_crimes | #22 / 32 | ↓28.7 | Source ↗official | |
| AnimalHarmBenchscore | #1 / 10 | ↑0.068 | Source ↗official | |
| BlueBench AttaQ-100attaq_harmlessness_reward_pct | #2 / 18 | ↑88.04 | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #79 / 104 | ↓63.06 | Source ↗official | |
| OR-Benchover_refusal_rate | #3 / 25 | ↓9.8 | Source ↗official | |
| OR-Benchtoxic_acceptance_rate | #25 / 25 | ↓27.2 | 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.
| Eval / sub-eval | Rank | Value | Distribution | Source |
|---|---|---|---|---|
| Adversarial Robustnessscore | #2 / 8 | ↑37 | Source ↗official | |
| AILuminate General Purpose AI Chatchild_sexual_exploitation | #16 / 32 | ↑21.5 | Source ↗official | |
| AILuminate General Purpose AI Chatdefamation | #9 / 32 | ↑26.3 | Source ↗official | |
| AILuminate General Purpose AI Chathate | #15 / 32 | ↑9.1 | Source ↗official | |
| AILuminate General Purpose AI Chatindiscriminate_weapons_cbrne | #11 / 32 | ↑31.9 | Source ↗official | |
| AILuminate General Purpose AI Chatintellectual_property | #12 / 32 | ↑18 | Source ↗official | |
| AILuminate General Purpose AI Chatnon_violent_crimes | #7 / 32 | ↑31.1 | Source ↗official | |
| AILuminate General Purpose AI Chatprivacy | #8 / 32 | ↑25.9 | Source ↗official | |
| AILuminate General Purpose AI Chatsex_related_crimes | #10 / 32 | ↑21.2 | Source ↗official | |
| AILuminate General Purpose AI Chatsexual_content | #11 / 32 | ↑34.3 | Source ↗official | |
| AILuminate General Purpose AI Chatspecialized_advice | #10 / 32 | ↑22.7 | Source ↗official | |
| AILuminate General Purpose AI Chatsuicide_self_harm | #19 / 32 | ↑12 | Source ↗official | |
| AILuminate General Purpose AI Chatviolent_crimes | #11 / 32 | ↑28.7 | Source ↗official | |
| BlueBench AttaQ-100attaq_harmlessness_reward_pct | #17 / 18 | ↓88.04 | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #26 / 104 | ↑63.06 | Source ↗official | |
| OR-Benchover_refusal_rate | #3 / 25 | ↓9.8 | Source ↗official | |
| OR-Benchtoxic_acceptance_rate | #1 / 25 | ↑27.2 | Source ↗official |