Model profile
Llama 4 Scout
Evidence summary
Llama 4 Scout has an estimated overall rank of #256; its 90% source-sensitivity interval is #190–#267. Its behavior-only rank is #257; company governance moves the combined estimate to #256. Published evidence spans 19 evals and 7 of 7 behavior components. Its strongest relative result is PHARE (bias_resistance_diagnostic, #7 of 66); its weakest is JuICE Cultural-Error Span Detection (f1, #10 of 10).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗llama-4-scout
- OpenRouter ↗meta-llama/llama-4-scout
- Official model page ↗Family-level model document · meta · first party
- Release source ↗direct research preview identity date
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 | #147 / 330 | ↓0.7935 | Source ↗official | |
| AgentDrive Safety Compliancescr | #36 / 48 | ↑58.75 | Source ↗official | |
| AIRBench 2024 Safety Scenariossafety_scenarios | #63 / 80 | ↑0.523 | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #68 / 106 | ↑0.19 | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #74 / 104 | ↓58.72 | Source ↗official | |
| DystopiaBenchbasaglia_score | #34 / 50 | ↓67.87 | Source ↗official | |
| DystopiaBenchbaudrillard_score | #45 / 50 | ↓74.3 | Source ↗official | |
| DystopiaBenchhuxley_score | #24 / 50 | ↓70.13 | Source ↗official | |
| DystopiaBenchlaguardia_score | #29 / 50 | ↓67.9 | Source ↗official | |
| DystopiaBenchorwell_score | #23 / 50 | ↓68.6 | Source ↗official | |
| DystopiaBenchpetrov_score | #20 / 50 | ↓70.47 | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #145 / 241 | ↑13.95 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #127 / 241 | ↑87.83 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #123 / 241 | ↑68.33 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #69 / 239 | ↑98.05 | Source ↗official | |
| HELM Safetyanthropic_red_team | #67 / 80 | ↑0.965 | Source ↗official | |
| HELM Safetybbq | #60 / 80 | ↑0.875 | Source ↗official | |
| HELM Safetyharmbench | #62 / 80 | ↑0.6 | Source ↗official | |
| HELM Safetysimple_safety_tests | #58 / 80 | ↑0.97 | Source ↗official | |
| HELM Safetyxstest | #43 / 80 | ↑0.956 | Source ↗official | |
| JuICE Cultural-Error Span Detectionf1 | #10 / 10 | ↑0.2805 | Source ↗official | |
| MT-JailBench CrescendoXsafety_score | #20 / 21 | ↑10.06 | Source ↗official | |
| PHAREbias_resistance_diagnostic | #7 / 66 | ↑0.671 | Source ↗official | |
| PHAREhallucination_resistance_diagnostic | #69 / 70 | ↑0.582 | Source ↗official | |
| PHAREharm_resistance_diagnostic | #65 / 70 | ↑0.8104 | Source ↗official | |
| PHAREjailbreak_resistance_diagnostic | #35 / 67 | ↑0.4893 | Source ↗official | |
| SOSBenchbiology_pvr | #13 / 23 | ↓0.488 | Source ↗official | |
| SOSBenchchemistry_pvr | #12 / 23 | ↓0.436 | Source ↗official | |
| SOSBenchmedicine_pvr | #13 / 23 | ↓0.688 | Source ↗official | |
| SOSBenchpharmacology_pvr | #14 / 23 | ↓0.874 | Source ↗official | |
| SOSBenchphysics_pvr | #13 / 23 | ↓0.492 | Source ↗official | |
| SOSBenchpsychology_pvr | #13 / 23 | ↓0.51 | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #22 / 113 | ↑6.97 | Source ↗official | |
| SpeciEvalland_animal_4ns | #103 / 113 | ↓5 | Source ↗official | |
| SpeciEvalsea_animal_4ns | #108 / 113 | ↓5.33 | Source ↗official | |
| SpeciEvalspeciesism | #58 / 113 | ↓2.02 | Source ↗official | |
| UAVBench safety-critical decision recognitionethical_safety_critical_accuracy | #19 / 27 | ↑0.635 | Source ↗official | |
| Vectara HHEM Factual Consistencyfactual_consistency_rate | #34 / 94 | ↑92.3 | Source ↗official | |
| Vigil Mental Health Safetyoverall_score | #23 / 23 | ↑20 | 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.