← Models

Model profile

Mistral Large 3

Mistral AIdeveloper
2025-12-01release date
#321 / 346Safety rank
#2 / 662Freedom rank

Evidence summary

Safety. Mistral Large 3 has an estimated Safety rank of #321; its 90% source-sensitivity interval is #184–#335. Its behavior-only rank is #313; company governance moves the combined estimate to #321. Published Safety evidence spans 16 eval lineages and 6 of 7 components. Its strongest relative result is HUMAINE Trust, Ethics and Safety (trust_ethics_safety_score, #3 of 54); its weakest is Adversarial Humanities Benchmark (AHB) — Table 5 (non_violent_crimes, #31 of 31).

Freedom. Mistral Large 3 has an estimated Freedom rank of #2; its 90% source-sensitivity interval is #5–#70. Published Freedom evidence spans 6 eval lineages and 1 of 1 components. Its strongest relative result is PHARE (jailbreak_resistance_diagnostic, #1 of 67); its weakest is Adversarial Humanities Benchmark (AHB) — Table 5 (privacy, #12 of 31).

Compare this model

Only models sharing at least one published sub-eval are listed.

Official and reference links

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-evalRankValueDistributionSource
AA-Omnisciencehallucination_rate#240 / 358↓0.8596Source ↗official
Adversarial Humanities Benchmark (AHB) — Table 5child_sexual_exploitation#30 / 31↓80Source ↗official
Adversarial Humanities Benchmark (AHB) — Table 5defamation#23 / 31↓79.1Source ↗official
Adversarial Humanities Benchmark (AHB) — Table 5hate#29 / 31↓73.3Source ↗official
Adversarial Humanities Benchmark (AHB) — Table 5indiscriminate_weapons_cbrne#30 / 31↓90.5Source ↗official
Adversarial Humanities Benchmark (AHB) — Table 5intellectual_property#27 / 31↓72.7Source ↗official
Adversarial Humanities Benchmark (AHB) — Table 5non_violent_crimes#31 / 31↓97.7Source ↗official
Adversarial Humanities Benchmark (AHB) — Table 5privacy#20 / 31↓76.1Source ↗official
Adversarial Humanities Benchmark (AHB) — Table 5sex_related_crimes#28 / 31↓86.4Source ↗official
Adversarial Humanities Benchmark (AHB) — Table 5sexual_content#29 / 31↓67.4Source ↗official
Adversarial Humanities Benchmark (AHB) — Table 5specialized_advice#31 / 31↓77.1Source ↗official
Adversarial Humanities Benchmark (AHB) — Table 5suicide_self_harm#31 / 31↓84.4Source ↗official
Adversarial Humanities Benchmark (AHB) — Table 5violent_crimes#25 / 31↓83.3Source ↗official
Alignment Leaderboardcorrigibility#24 / 24↑3.845Source ↗official
Alignment Leaderboardhonesty#23 / 24↑3.083Source ↗official
Alignment Leaderboardnon_manipulation#24 / 24↑2.402Source ↗official
Alignment Leaderboardrobustness#24 / 24↑2.28Source ↗official
Alignment Leaderboardsafety#24 / 24↑2.865Source ↗official
Alignment Leaderboardscheming#24 / 24↑3.211Source ↗official
Arena Factuality — Text Arena (factuality-only weighting)factuality_bt_rating#79 / 111↑1422.0Source ↗official
AuAu Authoritarian Response Auditbehavioral_vignette_arr_percent#14 / 17↓28Source ↗official
AuAu Authoritarian Response Auditrealistic_prompt_arr_percent#12 / 17↓2.7Source ↗official
BioTIERpermit_compliance_pct#8 / 52↑99.6Source ↗official
BioTIERrefuse_compliance_pct#49 / 52↑9.5Source ↗official
BullshitBench v2clear_pushback_rate#121 / 122↑0.02Source ↗official
Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate#101 / 104↓80.16Source ↗official
DystopiaBenchbasaglia_score#41 / 50↓69.8Source ↗official
DystopiaBenchbaudrillard_score#47 / 50↓77.73Source ↗official
DystopiaBenchhuxley_score#38 / 50↓75.8Source ↗official
DystopiaBenchlaguardia_score#36 / 50↓69Source ↗official
DystopiaBenchorwell_score#29 / 50↓70.93Source ↗official
DystopiaBenchpetrov_score#41 / 50↓78.47Source ↗official
HUMAINE Trust, Ethics and Safetytrust_ethics_safety_score#3 / 54↑29.36Source ↗official
LiveSecBenchethics#34 / 43↑24.59Source ↗official
LiveSecBenchfactuality#24 / 43↑43.2Source ↗official
LiveSecBenchlegality#33 / 43↑20.79Source ↗official
LiveSecBenchprivacy#41 / 43↑10.14Source ↗official
LiveSecBenchpsychological_health#28 / 43↑43.44Source ↗official
PHAREbias_resistance_diagnostic#10 / 66↑0.6272Source ↗official
PHAREhallucination_resistance_diagnostic#51 / 70↑0.6803Source ↗official
PHAREharm_resistance_diagnostic#54 / 70↑0.8806Source ↗official
PHAREjailbreak_resistance_diagnostic#67 / 67↑0.249Source ↗official
RealityTest — Text AI-Identity Disclosuredisclosure_probability#17 / 17↑0.078Source ↗official
RefusalBenchyouden_j#18 / 19↑0.004237Source ↗official
StereoTales Harmful Associationsbenign_significant_association_score#10 / 23↑86.47Source ↗official
Vectara HHEM Factual Consistencyfactual_consistency_rate#83 / 94↑85.5Source ↗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-evalRankValueDistributionSource
Adversarial Humanities Benchmark (AHB) — Table 5child_sexual_exploitation#2 / 31↑80Source ↗official
Adversarial Humanities Benchmark (AHB) — Table 5defamation#9 / 31↑79.1Source ↗official
Adversarial Humanities Benchmark (AHB) — Table 5hate#3 / 31↑73.3Source ↗official
Adversarial Humanities Benchmark (AHB) — Table 5indiscriminate_weapons_cbrne#2 / 31↑90.5Source ↗official
Adversarial Humanities Benchmark (AHB) — Table 5intellectual_property#5 / 31↑72.7Source ↗official
Adversarial Humanities Benchmark (AHB) — Table 5non_violent_crimes#1 / 31↑97.7Source ↗official
Adversarial Humanities Benchmark (AHB) — Table 5privacy#12 / 31↑76.1Source ↗official
Adversarial Humanities Benchmark (AHB) — Table 5sex_related_crimes#4 / 31↑86.4Source ↗official
Adversarial Humanities Benchmark (AHB) — Table 5sexual_content#3 / 31↑67.4Source ↗official
Adversarial Humanities Benchmark (AHB) — Table 5specialized_advice#1 / 31↑77.1Source ↗official
Adversarial Humanities Benchmark (AHB) — Table 5suicide_self_harm#1 / 31↑84.4Source ↗official
Adversarial Humanities Benchmark (AHB) — Table 5violent_crimes#7 / 31↑83.3Source ↗official
BioTIERpermit_compliance_pct#8 / 52↑99.6Source ↗official
BioTIERrefuse_compliance_pct#4 / 52↓9.5Source ↗official
Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate#4 / 104↑80.16Source ↗official
LiveSecBenchethics#10 / 43↓24.59Source ↗official
LiveSecBenchlegality#10 / 43↓20.79Source ↗official
LiveSecBenchprivacy#3 / 43↓10.14Source ↗official
LiveSecBenchpsychological_health#16 / 43↓43.44Source ↗official
PHAREharm_resistance_diagnostic#17 / 70↓0.8806Source ↗official
PHAREjailbreak_resistance_diagnostic#1 / 67↓0.249Source ↗official
UGI Leaderboard — base-model willingnesswillingness_adherence_score#4 / 156↑8.25Source ↗official
UGI Leaderboard — base-model willingnesswillingness_direct_score#21 / 156↑5Source ↗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.

UGI Political Values

DimensionValueDistribution
Political Lean-11.2
Government45.1
Diplomacy61.8
Economy43.7
Society59.8

Taiwan Sovereignty Benchmark Pro

DimensionValueDistribution
Pro-Taiwan rubric compatibility45
Warning-phrase rate30
Soft-censorship rate5
API-error rate0

The Economist World Values Survey Cultural Map

DimensionValueDistribution
Survival ↔ Self-expression0.866
Traditional ↔ Secular0.9