Model profile
Claude Opus 4.5
Anthropicdeveloper
2025-11-24release date
#23 / 312overall rank
25eval lineages
Evidence summary
Claude Opus 4.5 has an estimated overall rank of #23; its 90% source-sensitivity interval is #8–#115. Its behavior-only rank is #28; company governance moves the combined estimate to #23. Published evidence spans 25 evals and 7 of 7 behavior components. Its strongest relative result is Cisco AI Defense Rolling Single-Turn Leaderboard (single_turn_attack_success_rate, #1 of 104); its weakest is ANIMA (score, #22 of 22).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗claude-opus-4-5
- OpenRouter ↗anthropic/claude-opus-4.5
- System card ↗Exact 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 | Source |
|---|---|---|---|---|
| AA-Omnisciencehallucination_rate | #79 / 330 | ↓0.61 | Source ↗official | |
| Alignment Leaderboardcorrigibility | #2 / 24 | ↑4.563 | Source ↗official | |
| Alignment Leaderboardhonesty | #2 / 24 | ↑4.567 | Source ↗official | |
| Alignment Leaderboardnon_manipulation | #2 / 24 | ↑4.837 | Source ↗official | |
| Alignment Leaderboardrobustness | #7 / 24 | ↑3.907 | Source ↗official | |
| Alignment Leaderboardsafety | #2 / 24 | ↑4.788 | Source ↗official | |
| Alignment Leaderboardscheming | #1 / 24 | ↑4.801 | Source ↗official | |
| ANIMAscore | #22 / 22 | ↑0.5078 | Source ↗official | |
| Anthropic Claude Opus 4.5 System Cardagentic_coding_safety | #1 / 4 | ↑1 | Source ↗official | |
| Anthropic Claude Opus 4.5 System Cardclaude_code_malicious_refusal | #1 / 4 | ↑0.778 | Source ↗official | |
| Anthropic Claude Opus 4.5 System Cardcomputer_use_malicious_refusal | #1 / 4 | ↑0.8839 | Source ↗official | |
| Anthropic Claude Opus 4.5 System Cardharmful_request_safety | #1 / 2 | ↑0.9978 | Source ↗official | |
| Arena Factuality — Search Arena (factuality-only weighting)factuality_bt_rating | #13 / 30 | ↑1191.0 | Source ↗official | |
| Arena Factuality — Text Arena (factuality-only weighting)factuality_bt_rating | #19 / 112 | ↑1460.0 | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #4 / 106 | ↑0.845 | Source ↗official | |
| CAIS Risk Indexagent_red_teaming | #5 / 45 | ↓42.5 | Source ↗official | |
| CAIS Risk Indexbioweapons_assistance | #9 / 50 | ↓35.3 | Source ↗official | |
| CAIS Risk Indexhle_overconfidence | #28 / 51 | ↓56.9 | Source ↗official | |
| CAIS Risk Indexmachiavelli | #4 / 47 | ↓80.8 | Source ↗official | |
| CAIS Risk Indexmask | #5 / 53 | ↓5.9 | Source ↗official | |
| CAIS Risk Indexpolitical_manipulation | #20 / 34 | ↓49.4 | Source ↗official | |
| CAIS Risk Indextextquests_harm | #30 / 50 | ↓20.3 | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #1 / 104 | ↓2.193 | Source ↗official | |
| Claude Sonnet 4.6 Overrefusalhigher_difficulty_overrefusal_rate | #3 / 5 | ↓0.83 | Source ↗official | |
| Claude Sonnet 4.6 Overrefusaloverall_overrefusal_rate | #5 / 5 | ↓0.8 | Source ↗official | |
| Constitutional Following — Anthropic Constitutionconstitutional_following_score | #3 / 7 | ↑95.6 | Source ↗official | |
| FORTRESSaverage_risk_score | #2 / 49 | ↓11.59 | Source ↗official | |
| FORTRESSover_refusal_score | #30 / 48 | ↓5.145 | Source ↗official | |
| HUMAINE Trust, Ethics and Safetytrust_ethics_safety_score | #40 / 54 | ↑25.54 | Source ↗official | |
| Humanity's Last Exam RMS calibration error (Scale Labs)calibrationError | #11 / 39 | ↓55.5 | Source ↗official | |
| MACHIAVELLIdeception_relative_random_pct | #4 / 50 | ↓80.8 | Source ↗official | |
| MASKlying_probability_pct | #7 / 53 | ↓10.17 | Source ↗official | |
| MonitoringBench Full-Trajectory Monitorfull_trajectory_catch_rate_at_1pct_fpr_percent | #1 / 13 | ↑69.9 | Source ↗official | |
| PHAREbias_resistance_diagnostic | #9 / 66 | ↑0.632 | Source ↗official | |
| PHAREhallucination_resistance_diagnostic | #1 / 70 | ↑0.8823 | Source ↗official | |
| PHAREharm_resistance_diagnostic | #8 / 70 | ↑0.9825 | Source ↗official | |
| PHAREjailbreak_resistance_diagnostic | #3 / 67 | ↑0.7983 | Source ↗official | |
| RefusalBenchyouden_j | #3 / 19 | ↑0.6709 | Source ↗official | |
| Reward Hacking Benchmarkintegrity_score | #1 / 13 | ↑100 | Source ↗official | |
| SM-Benchadversarial | #26 / 79 | ↑83.9 | Source ↗official | |
| SM-Benchambiguous_interpretation | #18 / 79 | ↑89.58 | Source ↗official | |
| SM-Benchanti_hallucination | #8 / 79 | ↑98.95 | Source ↗official | |
| SM-Bencheq_boundaries | #60 / 79 | ↑56.46 | Source ↗official | |
| SM-Benchoverfit | #14 / 79 | ↑88.52 | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #75 / 113 | ↑6.65 | Source ↗official | |
| SpeciEvalland_animal_4ns | #105 / 113 | ↓5.03 | Source ↗official | |
| SpeciEvalsea_animal_4ns | #68 / 113 | ↓4.83 | Source ↗official | |
| SpeciEvalspeciesism | #101 / 113 | ↓2.75 | Source ↗official | |
| Vectara HHEM Factual Consistencyfactual_consistency_rate | #63 / 94 | ↑89.1 | 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.