Model profile
Claude 3 Sonnet
Anthropicdeveloper
2024-03-04release date
#196 / 267overall rank
13eval lineages
Evidence summary
Claude 3 Sonnet has an estimated overall rank of #196; its 90% source-sensitivity interval is #142–#205. Its behavior-only rank is #210; company governance moves the combined estimate to #196. Published evidence spans 13 evals and 6 of 7 behavior components. Its strongest relative result is HELM Safety (simple_safety_tests, #1 of 80); its weakest is HELM Safety (xstest, #74 of 80).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗claude-3-sonnet
- OpenRouter ↗anthropic/claude-3-sonnet
- System card ↗Family-level model document · Anthropic · 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 | Better | Source |
|---|---|---|---|---|---|
| AgentDojotargeted_attack_success_rate | #12 / 15 | 0.2671 | ↓ lower | Source ↗official | |
| AgentDojoutility_under_attack | #12 / 15 | 0.3323 | ↑ higher | Source ↗official | |
| AgentHarmharm_score | #7 / 12 | 20.7 | ↓ lower | Source ↗official | |
| AIRBench 2024 Safety Scenariossafety_scenarios | #13 / 80 | 0.847 | ↑ higher | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #22 / 105 | 7.827 | ↓ lower | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #81 / 260 | 21.71 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #12 / 260 | 96.33 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #42 / 260 | 93.33 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #21 / 258 | 99.64 | ↑ higher | Source ↗official | |
| HELM Safetyanthropic_red_team | #7 / 80 | 0.998 | ↑ higher | Source ↗official | |
| HELM Safetybbq | #55 / 80 | 0.9 | ↑ higher | Source ↗official | |
| HELM Safetyharmbench | #15 / 80 | 0.958 | ↑ higher | Source ↗official | |
| HELM Safetysimple_safety_tests | #1 / 80 | 1 | ↑ higher | Source ↗official | |
| HELM Safetyxstest | #74 / 80 | 0.858 | ↑ higher | Source ↗official | |
| Large-scale Moral Machine experiment on LLMshuman_choice_distance | #30 / 39 | 1.249 | ↓ lower | Source ↗official | |
| OR-Benchover_refusal_rate | #22 / 25 | 94.5 | ↓ lower | Source ↗official | |
| OR-Benchtoxic_acceptance_rate | #2 / 25 | 0.3 | ↓ lower | Source ↗official | |
| SORRY-Benchavg | #9 / 51 | 0.11 | ↓ lower | Source ↗official |