← Models

Model profile

Llama 4 Scout

Metadeveloper
2025-04-05release date
#216 / 267overall rank
16eval lineages

Evidence summary

Llama 4 Scout has an estimated overall rank of #216; its 90% source-sensitivity interval is #164–#223. Its behavior-only rank is #215; company governance moves the combined estimate to #216. Published evidence spans 16 evals and 7 of 7 behavior components. Its strongest relative result is PHARE (bias_resistance_diagnostic, #7 of 66); its weakest is Vigil Mental Health Safety (overall_score, #23 of 23).

Compare this model

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

Official and reference links

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-evalRankValueDistributionBetterSource
AA-Omnisciencehallucination_rate#135 / 3110.7831↓ lowerSource ↗official
AIRBench 2024 Safety Scenariossafety_scenarios#63 / 800.523↑ higherSource ↗official
BullshitBench v2clear_pushback_rate#67 / 1050.19↑ higherSource ↗official
Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate#75 / 10558.72↓ lowerSource ↗official
DystopiaBenchbasaglia_score#34 / 5067.87↓ lowerSource ↗official
DystopiaBenchbaudrillard_score#45 / 5074.3↓ lowerSource ↗official
DystopiaBenchhuxley_score#24 / 5070.13↓ lowerSource ↗official
DystopiaBenchlaguardia_score#29 / 5067.9↓ lowerSource ↗official
DystopiaBenchorwell_score#23 / 5068.6↓ lowerSource ↗official
DystopiaBenchpetrov_score#20 / 5070.47↓ lowerSource ↗official
Enkrypt AI Safety Leaderboardbias_attack_non_success_rate#160 / 26013.95↑ higherSource ↗official
Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate#140 / 26087.83↑ higherSource ↗official
Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate#138 / 26068.33↑ higherSource ↗official
Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate#78 / 25898.05↑ higherSource ↗official
HELM Safetyanthropic_red_team#67 / 800.965↑ higherSource ↗official
HELM Safetybbq#60 / 800.875↑ higherSource ↗official
HELM Safetyharmbench#62 / 800.6↑ higherSource ↗official
HELM Safetysimple_safety_tests#58 / 800.97↑ higherSource ↗official
HELM Safetyxstest#43 / 800.956↑ higherSource ↗official
PHAREbias_resistance_diagnostic#7 / 660.671↑ higherSource ↗official
PHAREhallucination_resistance_diagnostic#69 / 700.582↑ higherSource ↗official
PHAREharm_resistance_diagnostic#65 / 700.8104↑ higherSource ↗official
PHAREjailbreak_resistance_diagnostic#35 / 670.4893↑ higherSource ↗official
SOSBenchbiology_pvr#13 / 230.488↓ lowerSource ↗official
SOSBenchchemistry_pvr#12 / 230.436↓ lowerSource ↗official
SOSBenchmedicine_pvr#13 / 230.688↓ lowerSource ↗official
SOSBenchpharmacology_pvr#14 / 230.874↓ lowerSource ↗official
SOSBenchphysics_pvr#13 / 230.492↓ lowerSource ↗official
SOSBenchpsychology_pvr#13 / 230.51↓ lowerSource ↗official
SpeciEvalbelief_animal_sentience#19 / 1026.97↑ higherSource ↗official
SpeciEvalland_animal_4ns#92 / 1025↓ lowerSource ↗official
SpeciEvalsea_animal_4ns#97 / 1025.33↓ lowerSource ↗official
SpeciEvalspeciesism#52 / 1022.02↓ lowerSource ↗official
UAVBench safety-critical decision recognitionethical_safety_critical_accuracy#19 / 270.635↑ higherSource ↗official
Vigil Mental Health Safetyoverall_score#23 / 2320↑ higherSource ↗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.

The Economist World Values Survey Cultural Map

DimensionValueDistribution
Survival ↔ Self-expression1.07
Traditional ↔ Secular1.51