Model profile
Claude Opus 5.5
Anthropicdeveloper
2026-09-22release date
#5 / 346Safety rank
#610 / 662Freedom rank
1discovery sources
Evidence summary
Safety. Claude Opus 5.5 has an estimated Safety rank of #5; its 90% source-sensitivity interval is #1–#50. Its behavior-only rank is #8; company governance moves the combined estimate to #5. Published Safety evidence spans 12 eval lineages and 7 of 7 components. Its strongest relative result is Enkrypt AI Safety Leaderboard (bias_attack_non_success_rate, #5 of 270); its weakest is Claude Opus 5.5 card — safety and refusal calibration (overall_harmful_api_harmless_rate_pct, #4 of 4).
Freedom. Claude Opus 5.5 has an estimated Freedom rank of #610; its 90% source-sensitivity interval is #299–#652. Published Freedom evidence spans 4 eval lineages and 1 of 1 components. Its strongest relative result is SM-Bench (overfit, #21 of 92); its weakest is Claude Opus 5.5 card — safety and refusal calibration (multiturn_biological_weapons_api_appropriate_rate_pct, #3 of 3).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗claude-opus-5-5
- OpenRouter ↗anthropic/claude-opus-5.5
- Release source ↗curated outward evidence
Safety evals
Rank and direction are specific to the Safety portfolio. Black marks the observed result; the grey dot marks the value implied by the Safety rank. Values stay on each source’s native scale.
| Eval / sub-eval | Rank | Value | Distribution | Source |
|---|---|---|---|---|
| AA-Omnisciencehallucination_rate | #103 / 358 | ↓0.5861 | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #20 / 122 | ↑0.625 | Source ↗official | |
| CAIS Risk Indexpolitical_manipulation | #2 / 51 | ↓29.1 | Source ↗official | |
| CheatBench direct cheating propensitybiology_bioinformatics_cheating_rate_pct | #3 / 12 | ↓20 | Source ↗official | |
| CheatBench direct cheating propensityboard_games_cheating_rate_pct | #1 / 12 | ↓0 | Source ↗official | |
| CheatBench direct cheating propensitycreative_writing_cheating_rate_pct | #1 / 12 | ↓0 | Source ↗official | |
| CheatBench direct cheating propensityknowledge_work_cheating_rate_pct | #1 / 12 | ↓0 | Source ↗official | |
| CheatBench direct cheating propensitymathematical_research_cheating_rate_pct | #1 / 12 | ↓62.5 | Source ↗official | |
| CheatBench direct cheating propensitymenial_computation_cheating_rate_pct | #2 / 12 | ↓10 | Source ↗official | |
| CheatBench direct cheating propensitymultimodal_cheating_rate_pct | #1 / 12 | ↓10 | Source ↗official | |
| CheatBench direct cheating propensitysoftware_engineering_cheating_rate_pct | #1 / 11 | ↓0 | Source ↗official | |
| CheatBench direct cheating propensitysvg_competition_cheating_rate_pct | #1 / 12 | ↓0 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestyaccepting_unverifiable_authorization | #2 / 5 | ↓2.43 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestyapproval_gate_bypass | #1 / 5 | ↓1.52 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestybiological_weapons_uplift | #1 / 5 | ↓1.02 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestychemical_weapons_uplift | #1 / 5 | ↓1.02 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestycircumventing_constraints_or_controls | #1 / 5 | ↓1.58 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestycoherence_between_actions_and_views | #1 / 5 | ↑7.51 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestycompliance_with_deception_toward_user | #1 / 5 | ↓1.05 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestycomplying_with_prompt_injections | #3 / 5 | ↓1.32 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestyconcealed_git_changes_spontaneous_disclosure_rate_pct | #1 / 5 | ↑96.9 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestycontrolled_substance_synthesis_uplift | #1 / 5 | ↓1.04 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestycooperation_with_exfiltration_or_safeguard_tampering | #1 / 5 | ↓1.03 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestydangerous_actions_oblivious_to_real_setting | #1 / 5 | ↓1.09 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestydestructive_action_rate_pct | #1 / 6 | ↓21 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestydisallowed_cyberoffense | #1 / 5 | ↓1.09 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestyencouragement_of_user_delusion | #1 / 5 | ↓1.04 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestyevasiveness_controversial_topics | #4 / 5 | ↓1.12 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestyexplosive_weapons_uplift | #1 / 5 | ↓1.02 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestyfalse_completion_claims | #1 / 5 | ↓1.14 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestyfraud | #1 / 5 | ↓1.1 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestyharmful_sexual_content | #1 / 5 | ↓1.04 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestyharmful_system_prompt_compliance | #1 / 5 | ↓1.08 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestyignoring_explicit_constraints | #1 / 5 | ↓1.69 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestyimportant_omissions | #1 / 5 | ↓1.65 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestyinput_hallucination | #1 / 5 | ↓1.24 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestyinternalized_propaganda | #1 / 5 | ↓1.04 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestyleaked_answer_undisclosed_use_rate_pct | #2 / 5 | ↓12.1 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestymask_honesty_rate_pct | #4 / 5 | ↑87.4 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestymilitary_grade_weapons | #1 / 5 | ↓1.04 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestymotivated_reasoning | #1 / 5 | ↓1.91 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestyoverrefusal | #4 / 5 | ↓1.61 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestypackage_registry_unencouraged_real_publish_rate_pct | #1 / 3 | ↓33.33 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestyradiological_nuclear_weapons_uplift | #1 / 5 | ↓1 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestyreckless_tool_use | #1 / 5 | ↓1.82 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestysandbox_moderate_or_higher_attempt_rate_pct | #1 / 4 | ↓0 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestysecurity_shortcut_for_expedience | #1 / 5 | ↓2.04 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestyself_preservation | #1 / 5 | ↓1.01 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestyself_serving_bias | #1 / 5 | ↓1.05 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestysupporting_user_autonomy | #2 / 5 | ↑7.56 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestysycophancy | #1 / 5 | ↓1.6 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestyundermining_liberal_democracy | #1 / 5 | ↓1.17 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestyunfaithful_thinking | #1 / 5 | ↓1.54 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestyunprompted_boundary_probing | #1 / 5 | ↓1.1 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestyunsanctioned_sandbox_escape_attempts | #1 / 5 | ↓1.36 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestyunsanctioned_third_party_contact | #1 / 5 | ↓1.28 | Source ↗official | |
| Claude Opus 5.5 card — alignment and honestyuser_deception | #1 / 5 | ↓1.12 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationchild_api_multiturn_appropriate_rate_pct | #3 / 4 | ↑84 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationchild_claude_ai_multiturn_appropriate_rate_pct | #2 / 4 | ↑99 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationclaude_code_dual_use_benign_success_rate_pct | #1 / 4 | ↑99.8 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationclaude_code_malicious_refusal_rate_pct | #4 / 4 | ↑79.8 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationelection_api_harmless_rate_pct | #4 / 4 | ↑99.33 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationelection_api_multiturn_appropriate_rate_pct | #4 / 4 | ↑87 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationelection_api_overrefusal_rate_pct | #1 / 4 | ↓0 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationelection_claude_ai_harmless_rate_pct | #1 / 4 | ↑100 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationelection_claude_ai_multiturn_appropriate_rate_pct | #4 / 4 | ↑85 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationelection_claude_ai_overrefusal_rate_pct | #1 / 4 | ↓0 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmalicious_computer_use_refusal_rate_pct | #4 / 4 | ↑79.46 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmultiturn_biological_weapons_api_appropriate_rate_pct | #1 / 3 | ↑89 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmultiturn_biological_weapons_claude_ai_appropriate_rate_pct | #1 / 3 | ↑94 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmultiturn_cyberattacks_api_appropriate_rate_pct | #2 / 3 | ↑95 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmultiturn_cyberattacks_claude_ai_appropriate_rate_pct | #1 / 3 | ↑100 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmultiturn_deadly_weapons_api_appropriate_rate_pct | #2 / 3 | ↑82 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmultiturn_deadly_weapons_claude_ai_appropriate_rate_pct | #2 / 3 | ↑93 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmultiturn_hate_and_discrimination_api_appropriate_rate_pct | #3 / 3 | ↑96 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmultiturn_hate_and_discrimination_claude_ai_appropriate_rate_pct | #3 / 3 | ↑88 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmultiturn_influence_operations_api_appropriate_rate_pct | #2 / 3 | ↑62 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmultiturn_influence_operations_claude_ai_appropriate_rate_pct | #2 / 3 | ↑63 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmultiturn_romance_scams_api_appropriate_rate_pct | #2 / 3 | ↑94 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmultiturn_romance_scams_claude_ai_appropriate_rate_pct | #2 / 3 | ↑94 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmultiturn_tracking_and_surveillance_api_appropriate_rate_pct | #3 / 3 | ↑65 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmultiturn_tracking_and_surveillance_claude_ai_appropriate_rate_pct | #3 / 3 | ↑69 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmultiturn_violent_extremism_api_appropriate_rate_pct | #2 / 3 | ↑95 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmultiturn_violent_extremism_claude_ai_appropriate_rate_pct | #1 / 3 | ↑98 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationoverall_benign_api_overrefusal_rate_pct | #2 / 4 | ↓0.03 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationoverall_benign_claude_ai_overrefusal_rate_pct | #2 / 4 | ↓0.38 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationoverall_harmful_api_harmless_rate_pct | #4 / 4 | ↑94.5 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationoverall_harmful_claude_ai_harmless_rate_pct | #2 / 4 | ↑99.51 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationselfharm_api_multiturn_appropriate_rate_pct | #2 / 4 | ↑66 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationselfharm_claude_ai_multiturn_appropriate_rate_pct | #1 / 4 | ↑94 | Source ↗official | |
| Claude system cards — Gray Swan Q1+Q2 indirect prompt injection k=15attack_success_probability_k15_pct | #1 / 12 | ↓1 | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #5 / 270 | ↑74.42 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #90 / 270 | ↑89.83 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #24 / 270 | ↑98.33 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #117 / 268 | ↑96.91 | Source ↗official | |
| FORTRESSaverage_risk_score | #6 / 60 | ↓13.18 | Source ↗official | |
| FORTRESSover_refusal_score | #46 / 59 | ↓9.5 | Source ↗official | |
| MentalHealthBenchoverall_model_performance_pct | #3 / 17 | ↑52.38 | Source ↗official | |
| Olam Diplomacy — Broken Promise Ratebroken_promise_rate_pct | #14 / 19 | ↓17.4 | Source ↗official | |
| SM-Benchadversarial | #46 / 92 | ↑81.95 | Source ↗official | |
| SM-Benchambiguous_interpretation | #3 / 92 | ↑96.73 | Source ↗official | |
| SM-Benchanti_hallucination | #10 / 92 | ↑98.95 | Source ↗official | |
| SM-Bencheq_boundaries | #55 / 92 | ↑62.36 | Source ↗official | |
| SM-Benchoverfit | #21 / 92 | ↑89.07 | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #95 / 131 | ↑6.6 | Source ↗official | |
| SpeciEvalland_animal_4ns | #84 / 131 | ↓4.65 | Source ↗official | |
| SpeciEvalsea_animal_4ns | #69 / 131 | ↓4.75 | Source ↗official | |
| SpeciEvalspeciesism | #88 / 131 | ↓2.2 | Source ↗official |
Freedom evals
Rank and direction are specific to the Freedom portfolio. Black marks the observed result; the grey dot marks the value implied by the Freedom rank. Values stay on each source’s native scale.
| Eval / sub-eval | Rank | Value | Distribution | Source |
|---|---|---|---|---|
| Claude Opus 5.5 card — safety and refusal calibrationchild_api_multiturn_appropriate_rate_pct | #1 / 4 | ↓84 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationchild_claude_ai_multiturn_appropriate_rate_pct | #2 / 4 | ↓99 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationelection_api_harmless_rate_pct | #1 / 4 | ↓99.33 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationelection_api_multiturn_appropriate_rate_pct | #1 / 4 | ↓87 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationelection_api_overrefusal_rate_pct | #1 / 4 | ↓0 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationelection_claude_ai_harmless_rate_pct | #1 / 4 | ↓100 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationelection_claude_ai_multiturn_appropriate_rate_pct | #1 / 4 | ↓85 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationelection_claude_ai_overrefusal_rate_pct | #1 / 4 | ↓0 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmultiturn_biological_weapons_api_appropriate_rate_pct | #3 / 3 | ↓89 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmultiturn_biological_weapons_claude_ai_appropriate_rate_pct | #3 / 3 | ↓94 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmultiturn_cyberattacks_api_appropriate_rate_pct | #2 / 3 | ↓95 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmultiturn_cyberattacks_claude_ai_appropriate_rate_pct | #2 / 3 | ↓100 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmultiturn_deadly_weapons_api_appropriate_rate_pct | #2 / 3 | ↓82 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmultiturn_deadly_weapons_claude_ai_appropriate_rate_pct | #2 / 3 | ↓93 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmultiturn_hate_and_discrimination_api_appropriate_rate_pct | #1 / 3 | ↓96 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmultiturn_hate_and_discrimination_claude_ai_appropriate_rate_pct | #1 / 3 | ↓88 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmultiturn_influence_operations_api_appropriate_rate_pct | #2 / 3 | ↓62 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmultiturn_influence_operations_claude_ai_appropriate_rate_pct | #2 / 3 | ↓63 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmultiturn_romance_scams_api_appropriate_rate_pct | #2 / 3 | ↓94 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmultiturn_romance_scams_claude_ai_appropriate_rate_pct | #2 / 3 | ↓94 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmultiturn_tracking_and_surveillance_api_appropriate_rate_pct | #1 / 3 | ↓65 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmultiturn_tracking_and_surveillance_claude_ai_appropriate_rate_pct | #1 / 3 | ↓69 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmultiturn_violent_extremism_api_appropriate_rate_pct | #2 / 3 | ↓95 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationmultiturn_violent_extremism_claude_ai_appropriate_rate_pct | #3 / 3 | ↓98 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationoverall_benign_api_overrefusal_rate_pct | #2 / 4 | ↓0.03 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationoverall_benign_claude_ai_overrefusal_rate_pct | #2 / 4 | ↓0.38 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationoverall_harmful_api_harmless_rate_pct | #1 / 4 | ↓94.5 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationoverall_harmful_claude_ai_harmless_rate_pct | #3 / 4 | ↓99.51 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationselfharm_api_multiturn_appropriate_rate_pct | #3 / 4 | ↓66 | Source ↗official | |
| Claude Opus 5.5 card — safety and refusal calibrationselfharm_claude_ai_multiturn_appropriate_rate_pct | #3 / 4 | ↓94 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #178 / 270 | ↓89.83 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #241 / 270 | ↓98.33 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #148 / 268 | ↓96.91 | Source ↗official | |
| FORTRESSaverage_risk_score | #55 / 60 | ↑13.18 | Source ↗official | |
| FORTRESSover_refusal_score | #46 / 59 | ↓9.5 | Source ↗official | |
| SM-Benchadversarial | #45 / 92 | ↓81.95 | Source ↗official | |
| SM-Bencheq_boundaries | #55 / 92 | ↑62.36 | Source ↗official | |
| SM-Benchoverfit | #21 / 92 | ↑89.07 | Source ↗official |