Model profile
Claude Sonnet 4.5
Anthropicdeveloper
2025-09-29release date
#27 / 309overall rank
39eval lineages
Evidence summary
Claude Sonnet 4.5 has an estimated overall rank of #27; its 90% source-sensitivity interval is #22–#100. Its behavior-only rank is #31; company governance moves the combined estimate to #27. Published evidence spans 39 evals and 7 of 7 behavior components. Its strongest relative result is HELM Safety (bbq, #1 of 80); its weakest is Anthropic Claude Haiku 4.5 System Card (agentic_coding_safety, #3 of 3).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗claude-4-5-sonnet
- OpenRouter ↗anthropic/claude-sonnet-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 | #54 / 328 | ↓0.4908 | Source ↗official | |
| AgentDrive Safety Compliancescr | #16 / 48 | ↑90 | Source ↗official | |
| AIRBench 2024 Safety Scenariossafety_scenarios | #2 / 80 | ↑0.898 | Source ↗official | |
| Alignment Leaderboardcorrigibility | #5 / 24 | ↑4.485 | Source ↗official | |
| Alignment Leaderboardhonesty | #1 / 24 | ↑4.581 | Source ↗official | |
| Alignment Leaderboardnon_manipulation | #1 / 24 | ↑4.866 | Source ↗official | |
| Alignment Leaderboardrobustness | #5 / 24 | ↑4.027 | Source ↗official | |
| Alignment Leaderboardsafety | #1 / 24 | ↑4.885 | Source ↗official | |
| Alignment Leaderboardscheming | #2 / 24 | ↑4.735 | Source ↗official | |
| ANIMAscore | #21 / 22 | ↑0.5274 | Source ↗official | |
| Anthropic Claude Haiku 4.5 System Cardagentic_coding_safety | #3 / 3 | ↑0.987 | Source ↗official | |
| Anthropic Claude Haiku 4.5 System Cardclaude_code_malicious_refusal | #3 / 3 | ↑0.6694 | Source ↗official | |
| Anthropic Claude Opus 4.5 System Cardagentic_coding_safety | #3 / 4 | ↑0.987 | Source ↗official | |
| Anthropic Claude Opus 4.5 System Cardclaude_code_malicious_refusal | #3 / 4 | ↑0.6306 | Source ↗official | |
| Anthropic Claude Opus 4.5 System Cardcomputer_use_malicious_refusal | #2 / 4 | ↑0.8303 | Source ↗official | |
| Anthropic Claude Sonnet 4.5 System Cardagentic_coding_safety | #1 / 2 | ↑0.987 | Source ↗official | |
| Anthropic Claude Sonnet 4.5 System Cardbenign_request_refusal | #1 / 2 | ↓0.0002 | Source ↗official | |
| Anthropic Claude Sonnet 4.5 System Cardclaude_code_malicious_refusal | #1 / 2 | ↑0.9556 | Source ↗official | |
| Anthropic Claude Sonnet 4.5 System Cardharmful_request_safety | #1 / 2 | ↑0.9929 | Source ↗official | |
| Arena Factuality — Search Arena (factuality-only weighting)factuality_bt_rating | #19 / 30 | ↑1182.0 | Source ↗official | |
| Arena Factuality — Text Arena (factuality-only weighting)factuality_bt_rating | #34 / 112 | ↑1448.0 | Source ↗official | |
| AutoElicit Transferabilityelicitation_rate | #3 / 7 | ↓41 | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #7 / 106 | ↑0.765 | Source ↗official | |
| CAIS Risk Indexagent_red_teaming | #16 / 44 | ↓64.6 | Source ↗official | |
| CAIS Risk Indexbioweapons_assistance | #2 / 49 | ↓12 | Source ↗official | |
| CAIS Risk Indexhle_overconfidence | #33 / 50 | ↓64.1 | Source ↗official | |
| CAIS Risk Indexmachiavelli | #7 / 46 | ↓81.6 | Source ↗official | |
| CAIS Risk Indexmask | #2 / 52 | ↓3.9 | Source ↗official | |
| CAIS Risk Indextextquests_harm | #21 / 49 | ↓17.1 | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #6 / 104 | ↓2.592 | Source ↗official | |
| Claude Sonnet 4.6 Overrefusalhigher_difficulty_overrefusal_rate | #5 / 5 | ↓8.5 | Source ↗official | |
| Claude Sonnet 4.6 Overrefusaloverall_overrefusal_rate | #1 / 5 | ↓0.08 | Source ↗official | |
| Claude Sonnet 4.6 User Wellbeingchild_benign_refusal_rate | #2 / 4 | ↓0.08 | Source ↗official | |
| Claude Sonnet 4.6 User Wellbeingchild_multiturn_appropriate_rate | #1 / 4 | ↑98 | Source ↗official | |
| Claude Sonnet 4.6 User Wellbeingchild_violative_harmless_rate | #4 / 4 | ↑99.65 | Source ↗official | |
| Claude Sonnet 4.6 User Wellbeingselfharm_benign_refusal_rate | #1 / 4 | ↓0.01 | Source ↗official | |
| Claude Sonnet 4.6 User Wellbeingselfharm_harmless_rate | #4 / 4 | ↑98.93 | Source ↗official | |
| Claude Sonnet 4.6 User Wellbeingselfharm_multiturn_appropriate_rate | #4 / 4 | ↑78 | Source ↗official | |
| Constitutional Following — Anthropic Constitutionconstitutional_following_score | #4 / 7 | ↑92.7 | Source ↗official | |
| Contextual MoralChoicehuman_agreement | #3 / 22 | ↑0.5 | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #19 / 241 | ↑47.55 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #141 / 241 | ↑87 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #6 / 241 | ↑99.44 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #48 / 239 | ↑98.73 | Source ↗official | |
| FORTRESSaverage_risk_score | #6 / 49 | ↓14.88 | Source ↗official | |
| FORTRESSover_refusal_score | #29 / 46 | ↓5.25 | Source ↗official | |
| HELM Safetyanthropic_red_team | #54 / 80 | ↑0.982 | Source ↗official | |
| HELM Safetybbq | #1 / 80 | ↑0.989 | Source ↗official | |
| HELM Safetyharmbench | #18 / 80 | ↑0.92 | Source ↗official | |
| HELM Safetysimple_safety_tests | #1 / 80 | ↑1 | Source ↗official | |
| HELM Safetyxstest | #33 / 80 | ↑0.964 | Source ↗official | |
| HUMAINE Trust, Ethics and Safetytrust_ethics_safety_score | #38 / 54 | ↑25.74 | Source ↗official | |
| MACHIAVELLIdeception_relative_random_pct | #7 / 50 | ↓81.6 | Source ↗official | |
| MASKlying_probability_pct | #4 / 53 | ↓8.735 | Source ↗official | |
| MonitoringBench Full-Trajectory Monitorfull_trajectory_catch_rate_at_1pct_fpr_percent | #2 / 13 | ↑53.1 | Source ↗official | |
| MORUscore | #13 / 13 | ↑65.95 | Source ↗official | |
| MT-JailBench CrescendoXsafety_score | #2 / 21 | ↑75.47 | Source ↗official | |
| PHAREbias_resistance_diagnostic | #26 / 66 | ↑0.4914 | Source ↗official | |
| PHAREhallucination_resistance_diagnostic | #4 / 70 | ↑0.87 | Source ↗official | |
| PHAREharm_resistance_diagnostic | #5 / 70 | ↑0.9905 | Source ↗official | |
| PHAREjailbreak_resistance_diagnostic | #7 / 67 | ↑0.7523 | Source ↗official | |
| Reward Hacking Benchmarkintegrity_score | #1 / 13 | ↑100 | Source ↗official | |
| SM-Benchadversarial | #49 / 79 | ↑80 | Source ↗official | |
| SM-Benchambiguous_interpretation | #47 / 79 | ↑82.74 | Source ↗official | |
| SM-Benchanti_hallucination | #8 / 79 | ↑98.95 | Source ↗official | |
| SM-Bencheq_boundaries | #64 / 79 | ↑54.21 | Source ↗official | |
| SM-Benchoverfit | #20 / 79 | ↑83.06 | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #50 / 113 | ↑6.83 | Source ↗official | |
| SpeciEvalland_animal_4ns | #29 / 113 | ↓4.3 | Source ↗official | |
| SpeciEvalsea_animal_4ns | #36 / 113 | ↓4.65 | Source ↗official | |
| SpeciEvalspeciesism | #48 / 113 | ↓1.92 | Source ↗official | |
| ThaiSafetyBenchsafety_score | #2 / 18 | ↑90.25 | Source ↗official | |
| TrustLLM contemporary collapsed applicationtrustllm | #1 / 8 | ↑0.64 | Source ↗official | |
| Vectara HHEM Factual Consistencyfactual_consistency_rate | #71 / 94 | ↑88 | Source ↗official | |
| Vigil Mental Health Safetyoverall_score | #3 / 23 | ↑75 | 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.