Model profile
Llama 3.1 405B Instruct
Metadeveloper
2024-07-23release date
#166 / 267overall rank
22eval lineages
Evidence summary
Llama 3.1 405B Instruct has an estimated overall rank of #166; its 90% source-sensitivity interval is #96–#205. Its behavior-only rank is #151; company governance moves the combined estimate to #166. Published evidence spans 22 evals and 6 of 7 behavior components. Its strongest relative result is PHARE (bias_resistance_diagnostic, #3 of 66); 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 ↗llama-3-1-instruct-405b
- Hugging Face ↗meta-llama/llama-3.1-405B
- OpenRouter ↗meta-llama/llama-3.1-405b-instruct
- Official model page ↗Exact model document · Reviewed official Hugging Face owner · official repository
- 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 | Better | Source |
|---|---|---|---|---|---|
| AA-Omnisciencehallucination_rate | #56 / 311 | 0.5103 | ↓ lower | Source ↗official | |
| AbstentionBenchanswer_unknown_f1 | #11 / 20 | 0.87 | ↑ higher | Source ↗official | |
| AbstentionBenchfalse_premise_f1 | #7 / 20 | 0.7164 | ↑ higher | Source ↗official | |
| AbstentionBenchstale_f1 | #9 / 20 | 0.6419 | ↑ higher | Source ↗official | |
| AbstentionBenchsubjective_f1 | #10 / 20 | 0.7422 | ↑ higher | Source ↗official | |
| AbstentionBenchunderspecified_context_f1 | #3 / 20 | 0.7157 | ↑ higher | Source ↗official | |
| AbstentionBenchunderspecified_intent_f1 | #14 / 20 | 0.7024 | ↑ higher | Source ↗official | |
| Adversarial Robustnessscore | #2 / 8 | 10 | ↓ lower | Source ↗official | |
| Agent-SafetyBenchcompromise_availability | #14 / 16 | 19.6 | ↑ higher | Source ↗official | |
| Agent-SafetyBenchharmful_vulnerable_code | #6 / 16 | 40.4 | ↑ higher | Source ↗official | |
| Agent-SafetyBenchleak_sensitive_information | #12 / 16 | 25.2 | ↑ higher | Source ↗official | |
| Agent-SafetyBenchphysical_harm | #12 / 16 | 24.4 | ↑ higher | Source ↗official | |
| Agent-SafetyBenchproduce_unsafe_information | #4 / 16 | 98.8 | ↑ higher | Source ↗official | |
| Agent-SafetyBenchproperty_loss | #14 / 16 | 27.6 | ↑ higher | Source ↗official | |
| Agent-SafetyBenchspread_unsafe_information | #6 / 16 | 14.4 | ↑ higher | Source ↗official | |
| Agent-SafetyBenchviolate_law_ethics | #5 / 16 | 32.8 | ↑ higher | Source ↗official | |
| AgentHarmharm_score | #2 / 12 | 4.3 | ↓ lower | Source ↗official | |
| AILuminate General Purpose AI Chatchild_sexual_exploitation | #16 / 32 | 20.4 | ↓ lower | Source ↗official | |
| AILuminate General Purpose AI Chatdefamation | #21 / 32 | 20.6 | ↓ lower | Source ↗official | |
| AILuminate General Purpose AI Chathate | #24 / 32 | 14.5 | ↓ lower | Source ↗official | |
| AILuminate General Purpose AI Chatindiscriminate_weapons_cbrne | #21 / 32 | 29.5 | ↓ lower | Source ↗official | |
| AILuminate General Purpose AI Chatintellectual_property | #23 / 32 | 19 | ↓ lower | Source ↗official | |
| AILuminate General Purpose AI Chatnon_violent_crimes | #23 / 32 | 23.9 | ↓ lower | Source ↗official | |
| AILuminate General Purpose AI Chatprivacy | #20 / 32 | 17.9 | ↓ lower | Source ↗official | |
| AILuminate General Purpose AI Chatsex_related_crimes | #22 / 32 | 19.9 | ↓ lower | Source ↗official | |
| AILuminate General Purpose AI Chatsexual_content | #24 / 32 | 35.2 | ↓ lower | Source ↗official | |
| AILuminate General Purpose AI Chatspecialized_advice | #21 / 32 | 20.2 | ↓ lower | Source ↗official | |
| AILuminate General Purpose AI Chatsuicide_self_harm | #21 / 32 | 15.7 | ↓ lower | Source ↗official | |
| AILuminate General Purpose AI Chatviolent_crimes | #23 / 32 | 30.9 | ↓ lower | Source ↗official | |
| AIRBench 2024 Safety Scenariossafety_scenarios | #52 / 80 | 0.586 | ↑ higher | Source ↗official | |
| Alignment Leaderboardcorrigibility | #20 / 24 | 4.058 | ↑ higher | Source ↗official | |
| Alignment Leaderboardhonesty | #17 / 24 | 3.475 | ↑ higher | Source ↗official | |
| Alignment Leaderboardnon_manipulation | #20 / 24 | 3.034 | ↑ higher | Source ↗official | |
| Alignment Leaderboardrobustness | #15 / 24 | 3.36 | ↑ higher | Source ↗official | |
| Alignment Leaderboardsafety | #17 / 24 | 3.692 | ↑ higher | Source ↗official | |
| Alignment Leaderboardscheming | #17 / 24 | 3.53 | ↑ higher | Source ↗official | |
| BlueBench AttaQ-100attaq_harmlessness_reward_pct | #17 / 18 | 81.35 | ↑ higher | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #44 / 105 | 31.56 | ↓ lower | Source ↗official | |
| Confabulationsconfabulation_rate | #14 / 52 | 14.36 | ↓ lower | Source ↗official | |
| FORTRESSaverage_risk_score | #19 / 49 | 20.61 | ↓ lower | Source ↗official | |
| FORTRESSover_refusal_score | #32 / 46 | 6.03 | ↓ lower | Source ↗official | |
| HELM Safetyanthropic_red_team | #67 / 80 | 0.965 | ↑ higher | Source ↗official | |
| HELM Safetybbq | #33 / 80 | 0.945 | ↑ higher | Source ↗official | |
| HELM Safetyharmbench | #59 / 80 | 0.627 | ↑ higher | Source ↗official | |
| HELM Safetysimple_safety_tests | #40 / 80 | 0.988 | ↑ higher | Source ↗official | |
| HELM Safetyxstest | #41 / 80 | 0.959 | ↑ higher | Source ↗official | |
| MASKlying_probability_pct | #23 / 53 | 33.3 | ↓ lower | Source ↗official | |
| PandaBench JBB direct-request panelsafety_rate | #43 / 46 | 0.92 | ↑ higher | Source ↗official | |
| PHAREbias_resistance_diagnostic | #3 / 66 | 0.7523 | ↑ higher | Source ↗official | |
| PHAREhallucination_resistance_diagnostic | #55 / 70 | 0.674 | ↑ higher | Source ↗official | |
| PHAREharm_resistance_diagnostic | #57 / 70 | 0.8649 | ↑ higher | Source ↗official | |
| PHAREjailbreak_resistance_diagnostic | #5 / 67 | 0.7655 | ↑ higher | Source ↗official | |
| SORRY-Benchavg | #20 / 51 | 0.24 | ↓ lower | Source ↗official | |
| SOSBenchbiology_pvr | #14 / 23 | 0.59 | ↓ lower | Source ↗official | |
| SOSBenchchemistry_pvr | #13 / 23 | 0.468 | ↓ lower | Source ↗official | |
| SOSBenchmedicine_pvr | #14 / 23 | 0.69 | ↓ lower | Source ↗official | |
| SOSBenchpharmacology_pvr | #11 / 23 | 0.764 | ↓ lower | Source ↗official | |
| SOSBenchphysics_pvr | #11 / 23 | 0.444 | ↓ lower | Source ↗official | |
| SOSBenchpsychology_pvr | #14 / 23 | 0.568 | ↓ lower | 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.