Model profile
Claude Opus 5
Anthropicdeveloper
2026-07-24release date
#1 / 309overall rank
15eval lineages
1discovery sources
Evidence summary
Claude Opus 5 has an estimated overall rank of #1; its 90% source-sensitivity interval is #1–#27. Published evidence spans 15 evals and 6 of 7 behavior components. Its strongest relative result is SM-Bench (ambiguous_interpretation, #1 of 79); its weakest is Olam Social Poker — Social Lie Rate (social_lie_rate_per_10000_turns, #18 of 19).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗claude-opus-5
- OpenRouter ↗anthropic/claude-opus-5
- System card ↗Exact model document · Anthropic · first party
- Release source ↗curated outward evidence
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 | #73 / 328 | ↓0.5955 | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #11 / 106 | ↑0.715 | Source ↗official | |
| CAIS Risk Indexagent_red_teaming | #5 / 44 | ↓43.3 | Source ↗official | |
| CAIS Risk Indexbioweapons_assistance | #5 / 49 | ↓23.5 | Source ↗official | |
| CAIS Risk Indexhle_overconfidence | #1 / 50 | ↓23.6 | Source ↗official | |
| CAIS Risk Indexmachiavelli | #42 / 46 | ↓96.5 | Source ↗official | |
| CAIS Risk Indexmask | #6 / 52 | ↓6.6 | Source ↗official | |
| CAIS Risk Indexpolitical_manipulation | #5 / 33 | ↓36.7 | Source ↗official | |
| CAIS Risk Indextextquests_harm | #42 / 49 | ↓22.6 | Source ↗official | |
| Gray Swan indirect prompt injection (15 attempts)attack_success_probability_k15_pct | #1 / 13 | ↓2 | Source ↗official | |
| Manager Coercion Benchcoercion_ladder_depth | #2 / 31 | ↓3.5 | Source ↗official | |
| Manager Coercion Benchfabrication_rate | #1 / 13 | ↓0 | Source ↗official | |
| Olam Social Poker — Social Lie Ratesocial_lie_rate_per_10000_turns | #18 / 19 | ↓163 | Source ↗official | |
| SM-Benchadversarial | #17 / 79 | ↑86.34 | Source ↗official | |
| SM-Benchambiguous_interpretation | #1 / 79 | ↑97.62 | Source ↗official | |
| SM-Benchanti_hallucination | #8 / 79 | ↑98.95 | Source ↗official | |
| SM-Bencheq_boundaries | #23 / 79 | ↑68.82 | Source ↗official | |
| SM-Benchoverfit | #12 / 79 | ↑91.26 | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #88 / 113 | ↑6.5 | Source ↗official | |
| SpeciEvalland_animal_4ns | #32 / 113 | ↓4.35 | Source ↗official | |
| SpeciEvalsea_animal_4ns | #14 / 113 | ↓4.38 | Source ↗official | |
| SpeciEvalspeciesism | #48 / 113 | ↓1.92 | Source ↗official | |
| TACbase_welfare_rate | #2 / 76 | ↑59.62 | Source ↗official | |
| UK AISI active safety-research compromise continuationactive_compromise_continuation_rate_pct | #1 / 5 | ↓0.1 | Source ↗official |