Model profile
Claude Haiku 4.5
Anthropicdeveloper
2025-10-15release date
#17 / 267overall rank
32eval lineages
Evidence summary
Claude Haiku 4.5 has an estimated overall rank of #17; its 90% source-sensitivity interval is #11–#71. Its behavior-only rank is #24; company governance moves the combined estimate to #17. Published evidence spans 32 evals and 7 of 7 behavior components. Its strongest relative result is AIRBench 2024 Safety Scenarios (safety_scenarios, #1 of 80); its weakest is Anthropic Claude Haiku 4.5 System Card (harmful_request_safety, #2 of 2).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗c2b1e769-7aee-4669-8076-73918bdebf6c
- OpenRouter ↗anthropic/claude-haiku-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 | Better | Source |
|---|---|---|---|---|---|
| AA-Omnisciencehallucination_rate | #9 / 311 | 0.2471 | ↓ lower | Source ↗official | |
| AgentAbstainabstain | #5 / 17 | 65.6 | ↑ higher | Source ↗official | |
| AgentAbstaincar | #6 / 17 | 61.8 | ↑ higher | Source ↗official | |
| AgentAbstainpaired | #5 / 17 | 49.7 | ↑ higher | Source ↗official | |
| AIRBench 2024 Safety Scenariossafety_scenarios | #1 / 80 | 0.932 | ↑ higher | Source ↗official | |
| ANIMAscore | #14 / 19 | 0.5862 | ↑ higher | Source ↗official | |
| Anthropic Claude Haiku 4.5 System Cardagentic_coding_safety | #1 / 3 | 1 | ↑ higher | Source ↗official | |
| Anthropic Claude Haiku 4.5 System Cardclaude_code_malicious_refusal | #2 / 3 | 0.6939 | ↑ higher | Source ↗official | |
| Anthropic Claude Haiku 4.5 System Cardharmful_request_safety | #2 / 2 | 0.9938 | ↑ higher | Source ↗official | |
| Anthropic Claude Opus 4.5 System Cardagentic_coding_safety | #1 / 4 | 1 | ↑ higher | Source ↗official | |
| Anthropic Claude Opus 4.5 System Cardclaude_code_malicious_refusal | #2 / 4 | 0.6939 | ↑ higher | Source ↗official | |
| Anthropic Claude Opus 4.5 System Cardcomputer_use_malicious_refusal | #3 / 4 | 0.7232 | ↑ higher | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #8 / 105 | 0.74 | ↑ higher | Source ↗official | |
| CAIS Risk Indexagent_red_teaming | #18 / 43 | 69.7 | ↓ lower | Source ↗official | |
| CAIS Risk Indexbioweapons_assistance | #34 / 48 | 84.9 | ↓ lower | Source ↗official | |
| CAIS Risk Indexhle_overconfidence | #31 / 49 | 60.5 | ↓ lower | Source ↗official | |
| CAIS Risk Indexmachiavelli | #6 / 45 | 81.5 | ↓ lower | Source ↗official | |
| CAIS Risk Indexmask | #3 / 51 | 4.6 | ↓ lower | Source ↗official | |
| CAIS Risk Indexpolitical_manipulation | #31 / 32 | 62.2 | ↓ lower | Source ↗official | |
| CAIS Risk Indextextquests_harm | #18 / 48 | 16.6 | ↓ lower | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #12 / 105 | 3.539 | ↓ lower | Source ↗official | |
| Claude Sonnet 4.6 Overrefusalhigher_difficulty_overrefusal_rate | #4 / 5 | 6.01 | ↓ lower | Source ↗official | |
| Claude Sonnet 4.6 Overrefusaloverall_overrefusal_rate | #3 / 5 | 0.26 | ↓ lower | Source ↗official | |
| Claude Sonnet 4.6 User Wellbeingchild_benign_refusal_rate | #1 / 4 | 0.04 | ↓ lower | Source ↗official | |
| Claude Sonnet 4.6 User Wellbeingchild_multiturn_appropriate_rate | #2 / 4 | 96 | ↑ higher | Source ↗official | |
| Claude Sonnet 4.6 User Wellbeingchild_violative_harmless_rate | #3 / 4 | 99.88 | ↑ higher | Source ↗official | |
| Claude Sonnet 4.6 User Wellbeingselfharm_benign_refusal_rate | #2 / 4 | 0.03 | ↓ lower | Source ↗official | |
| Claude Sonnet 4.6 User Wellbeingselfharm_harmless_rate | #3 / 4 | 99.67 | ↑ higher | Source ↗official | |
| Claude Sonnet 4.6 User Wellbeingselfharm_multiturn_appropriate_rate | #2 / 4 | 90 | ↑ higher | Source ↗official | |
| Contextual MoralChoicehuman_agreement | #6 / 22 | 0.45 | ↑ higher | Source ↗official | |
| DystopiaBenchbasaglia_score | #3 / 50 | 25.3 | ↓ lower | Source ↗official | |
| DystopiaBenchbaudrillard_score | #2 / 50 | 22.73 | ↓ lower | Source ↗official | |
| DystopiaBenchhuxley_score | #6 / 50 | 19.07 | ↓ lower | Source ↗official | |
| DystopiaBenchlaguardia_score | #4 / 50 | 34.47 | ↓ lower | Source ↗official | |
| DystopiaBenchorwell_score | #2 / 50 | 21.63 | ↓ lower | Source ↗official | |
| DystopiaBenchpetrov_score | #1 / 50 | 19.8 | ↓ lower | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #43 / 260 | 35.14 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #109 / 260 | 89.5 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #17 / 260 | 98.89 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #51 / 258 | 98.91 | ↑ higher | Source ↗official | |
| HELM Safetyanthropic_red_team | #64 / 80 | 0.969 | ↑ higher | Source ↗official | |
| HELM Safetybbq | #44 / 80 | 0.928 | ↑ higher | Source ↗official | |
| HELM Safetyharmbench | #14 / 80 | 0.959 | ↑ higher | Source ↗official | |
| HELM Safetysimple_safety_tests | #40 / 80 | 0.988 | ↑ higher | Source ↗official | |
| HELM Safetyxstest | #60 / 80 | 0.932 | ↑ higher | Source ↗official | |
| Inkling-Small model card — FORTRESSbenign_answer_rate | #8 / 10 | 94.1 | ↑ higher | Source ↗official | |
| Inkling-Small model card — FORTRESSharmful_refusal_rate | #1 / 10 | 91.3 | ↑ higher | Source ↗official | |
| Inkling-Small model card — StrongREJECTsafety_rate | #6 / 10 | 98.6 | ↑ higher | Source ↗official | |
| LiveSecBenchethics | #4 / 43 | 88.65 | ↑ higher | Source ↗official | |
| LiveSecBenchfactuality | #3 / 43 | 82.68 | ↑ higher | Source ↗official | |
| LiveSecBenchlegality | #1 / 43 | 98.62 | ↑ higher | Source ↗official | |
| LiveSecBenchprivacy | #1 / 43 | 97.17 | ↑ higher | Source ↗official | |
| LiveSecBenchpsychological_health | #2 / 43 | 90.02 | ↑ higher | Source ↗official | |
| MACHIAVELLIdeception_relative_random_pct | #6 / 50 | 81.5 | ↓ lower | Source ↗official | |
| MORUscore | #7 / 13 | 74.02 | ↑ higher | Source ↗official | |
| PHAREbias_resistance_diagnostic | #6 / 66 | 0.7066 | ↑ higher | Source ↗official | |
| PHAREhallucination_resistance_diagnostic | #10 / 70 | 0.8356 | ↑ higher | Source ↗official | |
| PHAREharm_resistance_diagnostic | #3 / 70 | 0.9993 | ↑ higher | Source ↗official | |
| PHAREjailbreak_resistance_diagnostic | #4 / 67 | 0.7851 | ↑ higher | Source ↗official | |
| SM-Benchadversarial | #61 / 73 | 76.59 | ↑ higher | Source ↗official | |
| SM-Benchambiguous_interpretation | #29 / 73 | 86.61 | ↑ higher | Source ↗official | |
| SM-Benchanti_hallucination | #37 / 73 | 92.15 | ↑ higher | Source ↗official | |
| SM-Bencheq_boundaries | #71 / 73 | 41.85 | ↑ higher | Source ↗official | |
| SM-Benchoverfit | #22 / 73 | 81.97 | ↑ higher | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #95 / 102 | 6.32 | ↑ higher | Source ↗official | |
| SpeciEvalland_animal_4ns | #16 / 102 | 4.25 | ↓ lower | Source ↗official | |
| SpeciEvalsea_animal_4ns | #22 / 102 | 4.53 | ↓ lower | Source ↗official | |
| SpeciEvalspeciesism | #61 / 102 | 2.15 | ↓ lower | Source ↗official | |
| UAVBench safety-critical decision recognitionethical_safety_critical_accuracy | #17 / 27 | 0.67 | ↑ higher | Source ↗official | |
| Vigil Mental Health Safetyoverall_score | #4 / 23 | 74 | ↑ higher | 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.
UGI Political Values
Agent-ValueBench Moral Foundations (MFT08)
Agent-ValueBench HEXACO
| Dimension | Value | Distribution |
|---|---|---|
| Openness to experience | 5.4 | |
| Honesty-humility | 5.5 | |
| Extraversion | 5.7 | |
| Agreeableness | 7.5 | |
| Conscientiousness | 7.1 |