Evaluation profile
Claude Fable 5.1 / Mythos 5.1 card — safety and refusal calibration
51sub-evals
0.109%Safety weight
0.567%Freedom weight
5components
Weights below are portfolio-specific global index weights.
Model score (see each sub-eval direction)Predicted score
About this eval
Harmlessness, benign-request calibration, political even-handedness, election integrity, and malicious agent-use refusal.
Included in the behavior ranking.
Sub-evals
| Measure | Component | Direction by ranking | Safety weight | Freedom weight |
|---|---|---|---|---|
| child_api_harmless_rate_pctfable-mythos51-system-card/safety-calibration.csv:child_api_harmless_rate_pctMeasures child api harmless rate in the card-defined API or product-surface condition. | Safety: human_rights_systemic_harm:0.300|ordinary_harm_misuse_resistance:0.700 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: higher · Freedom: lower | 0.00152% | 0.0142% |
| child_api_multiturn_appropriate_rate_pctfable-mythos51-system-card/safety-calibration.csv:child_api_multiturn_appropriate_rate_pctMeasures child api multiturn appropriate rate in the card-defined API or product-surface condition. | Safety: human_rights_systemic_harm:0.300|ordinary_harm_misuse_resistance:0.700 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: higher · Freedom: lower | 0.00152% | 0.0142% |
| child_api_overrefusal_rate_pctfable-mythos51-system-card/safety-calibration.csv:child_api_overrefusal_rate_pctMeasures child api overrefusal rate in the card-defined API or product-surface condition. | Safety: benign_helpfulness_non_overrefusal:1.000 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: lower · Freedom: lower | 0.00364% | 0.0142% |
| child_claude_ai_harmless_rate_pctfable-mythos51-system-card/safety-calibration.csv:child_claude_ai_harmless_rate_pctMeasures child claude ai harmless rate in the card-defined API or product-surface condition. | Safety: human_rights_systemic_harm:0.300|ordinary_harm_misuse_resistance:0.700 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: higher · Freedom: lower | 0.00136% | 0.0127% |
| child_claude_ai_multiturn_appropriate_rate_pctfable-mythos51-system-card/safety-calibration.csv:child_claude_ai_multiturn_appropriate_rate_pctMeasures child claude ai multiturn appropriate rate in the card-defined API or product-surface condition. | Safety: human_rights_systemic_harm:0.300|ordinary_harm_misuse_resistance:0.700 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: higher · Freedom: lower | 0.00136% | 0.0127% |
| child_claude_ai_overrefusal_rate_pctfable-mythos51-system-card/safety-calibration.csv:child_claude_ai_overrefusal_rate_pctMeasures child claude ai overrefusal rate in the card-defined API or product-surface condition. | Safety: benign_helpfulness_non_overrefusal:1.000 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: lower · Freedom: lower | 0.00326% | 0.0127% |
| claude_code_dual_use_benign_success_rate_pctfable-mythos51-system-card/safety-calibration.csv:claude_code_dual_use_benign_success_rate_pctMeasures claude code dual use benign success rate in the card-defined API or product-surface condition. | Safety: benign_helpfulness_non_overrefusal:1.000anthropic-fable-mythos51-system-card | Safety: higher | 0.00326% | — |
| claude_code_malicious_refusal_rate_pctfable-mythos51-system-card/safety-calibration.csv:claude_code_malicious_refusal_rate_pctMeasures claude code malicious refusal rate in the card-defined API or product-surface condition. | Safety: ordinary_harm_misuse_resistance:0.700|responsible_agency_control:0.300anthropic-fable-mythos51-system-card | Safety: higher | 0.00104% | — |
| disordered_eating_api_harmless_rate_pctfable-mythos51-system-card/safety-calibration.csv:disordered_eating_api_harmless_rate_pctMeasures disordered eating api harmless rate in the card-defined API or product-surface condition. | Safety: human_rights_systemic_harm:0.300|ordinary_harm_misuse_resistance:0.700 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: higher · Freedom: lower | 0.00152% | 0.0142% |
| disordered_eating_api_overrefusal_rate_pctfable-mythos51-system-card/safety-calibration.csv:disordered_eating_api_overrefusal_rate_pctMeasures disordered eating api overrefusal rate in the card-defined API or product-surface condition. | Safety: benign_helpfulness_non_overrefusal:1.000 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: lower · Freedom: lower | 0.00364% | 0.0142% |
| disordered_eating_claude_ai_harmless_rate_pctfable-mythos51-system-card/safety-calibration.csv:disordered_eating_claude_ai_harmless_rate_pctMeasures disordered eating claude ai harmless rate in the card-defined API or product-surface condition. | Safety: human_rights_systemic_harm:0.300|ordinary_harm_misuse_resistance:0.700 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: higher · Freedom: lower | 0.00136% | 0.0127% |
| disordered_eating_claude_ai_overrefusal_rate_pctfable-mythos51-system-card/safety-calibration.csv:disordered_eating_claude_ai_overrefusal_rate_pctMeasures disordered eating claude ai overrefusal rate in the card-defined API or product-surface condition. | Safety: benign_helpfulness_non_overrefusal:1.000 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: lower · Freedom: lower | 0.00326% | 0.0127% |
| election_api_harmless_rate_pctfable-mythos51-system-card/safety-calibration.csv:election_api_harmless_rate_pctMeasures election api harmless rate in the card-defined API or product-surface condition. | Safety: human_rights_systemic_harm:0.700|ordinary_harm_misuse_resistance:0.300 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: higher · Freedom: lower | 0.00272% | 0.0142% |
| election_api_multiturn_appropriate_rate_pctfable-mythos51-system-card/safety-calibration.csv:election_api_multiturn_appropriate_rate_pctMeasures election api multiturn appropriate rate in the card-defined API or product-surface condition. | Safety: human_rights_systemic_harm:0.700|ordinary_harm_misuse_resistance:0.300 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: higher · Freedom: lower | 0.00272% | 0.0142% |
| election_api_overrefusal_rate_pctfable-mythos51-system-card/safety-calibration.csv:election_api_overrefusal_rate_pctMeasures election api overrefusal rate in the card-defined API or product-surface condition. | Safety: benign_helpfulness_non_overrefusal:1.000 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: lower · Freedom: lower | 0.00364% | 0.0142% |
| election_claude_ai_harmless_rate_pctfable-mythos51-system-card/safety-calibration.csv:election_claude_ai_harmless_rate_pctMeasures election claude ai harmless rate in the card-defined API or product-surface condition. | Safety: human_rights_systemic_harm:0.700|ordinary_harm_misuse_resistance:0.300 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: higher · Freedom: lower | 0.00244% | 0.0127% |
| election_claude_ai_multiturn_appropriate_rate_pctfable-mythos51-system-card/safety-calibration.csv:election_claude_ai_multiturn_appropriate_rate_pctMeasures election claude ai multiturn appropriate rate in the card-defined API or product-surface condition. | Safety: human_rights_systemic_harm:0.700|ordinary_harm_misuse_resistance:0.300 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: higher · Freedom: lower | 0.00244% | 0.0127% |
| election_claude_ai_overrefusal_rate_pctfable-mythos51-system-card/safety-calibration.csv:election_claude_ai_overrefusal_rate_pctMeasures election claude ai overrefusal rate in the card-defined API or product-surface condition. | Safety: benign_helpfulness_non_overrefusal:1.000 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: lower · Freedom: lower | 0.00326% | 0.0127% |
| malicious_computer_use_refusal_rate_pctfable-mythos51-system-card/safety-calibration.csv:malicious_computer_use_refusal_rate_pctMeasures malicious computer use refusal rate in the card-defined API or product-surface condition. | Safety: ordinary_harm_misuse_resistance:0.700|responsible_agency_control:0.300anthropic-fable-mythos51-system-card | Safety: higher | 0.00104% | — |
| multiturn_biological_weapons_api_appropriate_rate_pctfable-mythos51-system-card/safety-calibration.csv:multiturn_biological_weapons_api_appropriate_rate_pctMeasures whether the model responds appropriately across a simulated multi-turn biological weapons conversation. | Safety: ordinary_harm_misuse_resistance:1.000 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: higher · Freedom: lower | 0.000624% | 0.0142% |
| multiturn_biological_weapons_claude_ai_appropriate_rate_pctfable-mythos51-system-card/safety-calibration.csv:multiturn_biological_weapons_claude_ai_appropriate_rate_pctMeasures whether the model responds appropriately across a simulated multi-turn biological weapons conversation. | Safety: ordinary_harm_misuse_resistance:1.000 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: higher · Freedom: lower | 0.000558% | 0.0127% |
| multiturn_cyberattacks_api_appropriate_rate_pctfable-mythos51-system-card/safety-calibration.csv:multiturn_cyberattacks_api_appropriate_rate_pctMeasures whether the model responds appropriately across a simulated multi-turn cyberattacks conversation. | Safety: ordinary_harm_misuse_resistance:1.000 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: higher · Freedom: lower | 0.000624% | 0.0142% |
| multiturn_cyberattacks_claude_ai_appropriate_rate_pctfable-mythos51-system-card/safety-calibration.csv:multiturn_cyberattacks_claude_ai_appropriate_rate_pctMeasures whether the model responds appropriately across a simulated multi-turn cyberattacks conversation. | Safety: ordinary_harm_misuse_resistance:1.000 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: higher · Freedom: lower | 0.000558% | 0.0127% |
| multiturn_deadly_weapons_api_appropriate_rate_pctfable-mythos51-system-card/safety-calibration.csv:multiturn_deadly_weapons_api_appropriate_rate_pctMeasures whether the model responds appropriately across a simulated multi-turn deadly weapons conversation. | Safety: ordinary_harm_misuse_resistance:1.000 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: higher · Freedom: lower | 0.000624% | 0.0142% |
| multiturn_deadly_weapons_claude_ai_appropriate_rate_pctfable-mythos51-system-card/safety-calibration.csv:multiturn_deadly_weapons_claude_ai_appropriate_rate_pctMeasures whether the model responds appropriately across a simulated multi-turn deadly weapons conversation. | Safety: ordinary_harm_misuse_resistance:1.000 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: higher · Freedom: lower | 0.000558% | 0.0127% |
| multiturn_hate_and_discrimination_api_appropriate_rate_pctfable-mythos51-system-card/safety-calibration.csv:multiturn_hate_and_discrimination_api_appropriate_rate_pctMeasures whether the model responds appropriately across a simulated multi-turn hate and discrimination conversation. | Safety: human_rights_systemic_harm:0.300|fairness_nondiscrimination:0.500|ordinary_harm_misuse_resistance:0.200 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: higher · Freedom: lower | 0.00271% | 0.0142% |
| multiturn_hate_and_discrimination_claude_ai_appropriate_rate_pctfable-mythos51-system-card/safety-calibration.csv:multiturn_hate_and_discrimination_claude_ai_appropriate_rate_pctMeasures whether the model responds appropriately across a simulated multi-turn hate and discrimination conversation. | Safety: human_rights_systemic_harm:0.300|fairness_nondiscrimination:0.500|ordinary_harm_misuse_resistance:0.200 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: higher · Freedom: lower | 0.00242% | 0.0127% |
| multiturn_influence_operations_api_appropriate_rate_pctfable-mythos51-system-card/safety-calibration.csv:multiturn_influence_operations_api_appropriate_rate_pctMeasures whether the model responds appropriately across a simulated multi-turn influence operations conversation. | Safety: human_rights_systemic_harm:0.700|ordinary_harm_misuse_resistance:0.300 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: higher · Freedom: lower | 0.00272% | 0.0142% |
| multiturn_influence_operations_claude_ai_appropriate_rate_pctfable-mythos51-system-card/safety-calibration.csv:multiturn_influence_operations_claude_ai_appropriate_rate_pctMeasures whether the model responds appropriately across a simulated multi-turn influence operations conversation. | Safety: human_rights_systemic_harm:0.700|ordinary_harm_misuse_resistance:0.300 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: higher · Freedom: lower | 0.00244% | 0.0127% |
| multiturn_romance_scams_api_appropriate_rate_pctfable-mythos51-system-card/safety-calibration.csv:multiturn_romance_scams_api_appropriate_rate_pctMeasures whether the model responds appropriately across a simulated multi-turn romance scams conversation. | Safety: human_rights_systemic_harm:0.500|ordinary_harm_misuse_resistance:0.500 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: higher · Freedom: lower | 0.00212% | 0.0142% |
| multiturn_romance_scams_claude_ai_appropriate_rate_pctfable-mythos51-system-card/safety-calibration.csv:multiturn_romance_scams_claude_ai_appropriate_rate_pctMeasures whether the model responds appropriately across a simulated multi-turn romance scams conversation. | Safety: human_rights_systemic_harm:0.500|ordinary_harm_misuse_resistance:0.500 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: higher · Freedom: lower | 0.0019% | 0.0127% |
| multiturn_tracking_and_surveillance_api_appropriate_rate_pctfable-mythos51-system-card/safety-calibration.csv:multiturn_tracking_and_surveillance_api_appropriate_rate_pctMeasures whether the model responds appropriately across a simulated multi-turn tracking and surveillance conversation. | Safety: human_rights_systemic_harm:0.700|ordinary_harm_misuse_resistance:0.300 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: higher · Freedom: lower | 0.00272% | 0.0142% |
| multiturn_tracking_and_surveillance_claude_ai_appropriate_rate_pctfable-mythos51-system-card/safety-calibration.csv:multiturn_tracking_and_surveillance_claude_ai_appropriate_rate_pctMeasures whether the model responds appropriately across a simulated multi-turn tracking and surveillance conversation. | Safety: human_rights_systemic_harm:0.700|ordinary_harm_misuse_resistance:0.300 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: higher · Freedom: lower | 0.00244% | 0.0127% |
| multiturn_violent_extremism_api_appropriate_rate_pctfable-mythos51-system-card/safety-calibration.csv:multiturn_violent_extremism_api_appropriate_rate_pctMeasures whether the model responds appropriately across a simulated multi-turn violent extremism conversation. | Safety: human_rights_systemic_harm:0.500|ordinary_harm_misuse_resistance:0.500 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: higher · Freedom: lower | 0.00212% | 0.0142% |
| multiturn_violent_extremism_claude_ai_appropriate_rate_pctfable-mythos51-system-card/safety-calibration.csv:multiturn_violent_extremism_claude_ai_appropriate_rate_pctMeasures whether the model responds appropriately across a simulated multi-turn violent extremism conversation. | Safety: human_rights_systemic_harm:0.500|ordinary_harm_misuse_resistance:0.500 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: higher · Freedom: lower | 0.0019% | 0.0127% |
| overall_benign_api_overrefusal_rate_pctfable-mythos51-system-card/safety-calibration.csv:overall_benign_api_overrefusal_rate_pctMeasures overall benign api overrefusal rate in the card-defined API or product-surface condition. | Safety: benign_helpfulness_non_overrefusal:1.000 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: lower · Freedom: lower | 0.00364% | 0.0142% |
| overall_benign_claude_ai_overrefusal_rate_pctfable-mythos51-system-card/safety-calibration.csv:overall_benign_claude_ai_overrefusal_rate_pctMeasures overall benign claude ai overrefusal rate in the card-defined API or product-surface condition. | Safety: benign_helpfulness_non_overrefusal:1.000 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: lower · Freedom: lower | 0.00326% | 0.0127% |
| overall_harmful_api_harmless_rate_pctfable-mythos51-system-card/safety-calibration.csv:overall_harmful_api_harmless_rate_pctMeasures overall harmful api harmless rate in the card-defined API or product-surface condition. | Safety: ordinary_harm_misuse_resistance:1.000 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: higher · Freedom: lower | 0.000624% | 0.0142% |
| overall_harmful_claude_ai_harmless_rate_pctfable-mythos51-system-card/safety-calibration.csv:overall_harmful_claude_ai_harmless_rate_pctMeasures overall harmful claude ai harmless rate in the card-defined API or product-surface condition. | Safety: ordinary_harm_misuse_resistance:1.000 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: higher · Freedom: lower | 0.000558% | 0.0127% |
| political_api_refusal_rate_pctfable-mythos51-system-card/safety-calibration.csv:political_api_refusal_rate_pctMeasures political api refusal rate in the card-defined API or product-surface condition. | Safety: benign_helpfulness_non_overrefusal:1.000anthropic-fable-mythos51-system-card | Safety: lower | 0.00364% | — |
| political_claude_ai_refusal_rate_pctfable-mythos51-system-card/safety-calibration.csv:political_claude_ai_refusal_rate_pctMeasures political claude ai refusal rate in the card-defined API or product-surface condition. | Safety: benign_helpfulness_non_overrefusal:1.000anthropic-fable-mythos51-system-card | Safety: lower | 0.00326% | — |
| political_evenhandedness_api_pctfable-mythos51-system-card/safety-calibration.csv:political_evenhandedness_api_pctMeasures political evenhandedness api in the card-defined API or product-surface condition. | Safety: fairness_nondiscrimination:1.000anthropic-fable-mythos51-system-card | Safety: higher | 0.00239% | — |
| political_evenhandedness_claude_ai_pctfable-mythos51-system-card/safety-calibration.csv:political_evenhandedness_claude_ai_pctMeasures political evenhandedness claude ai in the card-defined API or product-surface condition. | Safety: fairness_nondiscrimination:1.000anthropic-fable-mythos51-system-card | Safety: higher | 0.00214% | — |
| political_opposing_perspectives_api_pctfable-mythos51-system-card/safety-calibration.csv:political_opposing_perspectives_api_pctMeasures political opposing perspectives api in the card-defined API or product-surface condition. | Safety: fairness_nondiscrimination:1.000anthropic-fable-mythos51-system-card | Safety: higher | 0.00239% | — |
| political_opposing_perspectives_claude_ai_pctfable-mythos51-system-card/safety-calibration.csv:political_opposing_perspectives_claude_ai_pctMeasures political opposing perspectives claude ai in the card-defined API or product-surface condition. | Safety: fairness_nondiscrimination:1.000anthropic-fable-mythos51-system-card | Safety: higher | 0.00214% | — |
| selfharm_api_harmless_rate_pctfable-mythos51-system-card/safety-calibration.csv:selfharm_api_harmless_rate_pctMeasures selfharm api harmless rate in the card-defined API or product-surface condition. | Safety: human_rights_systemic_harm:0.300|ordinary_harm_misuse_resistance:0.700 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: higher · Freedom: lower | 0.00152% | 0.0142% |
| selfharm_api_multiturn_appropriate_rate_pctfable-mythos51-system-card/safety-calibration.csv:selfharm_api_multiturn_appropriate_rate_pctMeasures selfharm api multiturn appropriate rate in the card-defined API or product-surface condition. | Safety: human_rights_systemic_harm:0.300|ordinary_harm_misuse_resistance:0.700 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: higher · Freedom: lower | 0.00152% | 0.0142% |
| selfharm_api_overrefusal_rate_pctfable-mythos51-system-card/safety-calibration.csv:selfharm_api_overrefusal_rate_pctMeasures selfharm api overrefusal rate in the card-defined API or product-surface condition. | Safety: benign_helpfulness_non_overrefusal:1.000 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: lower · Freedom: lower | 0.00364% | 0.0142% |
| selfharm_claude_ai_harmless_rate_pctfable-mythos51-system-card/safety-calibration.csv:selfharm_claude_ai_harmless_rate_pctMeasures selfharm claude ai harmless rate in the card-defined API or product-surface condition. | Safety: human_rights_systemic_harm:0.300|ordinary_harm_misuse_resistance:0.700 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: higher · Freedom: lower | 0.00136% | 0.0127% |
| selfharm_claude_ai_multiturn_appropriate_rate_pctfable-mythos51-system-card/safety-calibration.csv:selfharm_claude_ai_multiturn_appropriate_rate_pctMeasures selfharm claude ai multiturn appropriate rate in the card-defined API or product-surface condition. | Safety: human_rights_systemic_harm:0.300|ordinary_harm_misuse_resistance:0.700 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: higher · Freedom: lower | 0.00136% | 0.0127% |
| selfharm_claude_ai_overrefusal_rate_pctfable-mythos51-system-card/safety-calibration.csv:selfharm_claude_ai_overrefusal_rate_pctMeasures selfharm claude ai overrefusal rate in the card-defined API or product-surface condition. | Safety: benign_helpfulness_non_overrefusal:1.000 · Freedom: user_direction_following:1.000anthropic-fable-mythos51-system-card | Safety: lower · Freedom: lower | 0.00326% | 0.0127% |
child_api_harmless_rate_pct
Measures child api harmless rate in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-fable-5 | 100 | official | |
| 1 | claude-mythos-5 | 100 | official | |
| 1 | claude-opus-5 | 100 | official | |
| 4 | claude-sonnet-5 | 99.95 | official | |
| 5 | claude-mythos-5.1 | 99.9 | official |
child_api_multiturn_appropriate_rate_pct
Measures child api multiturn appropriate rate in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-mythos-5 | 89 | official | |
| 2 | claude-fable-5 | 88 | official | |
| 2 | claude-sonnet-5 | 88 | official | |
| 4 | claude-opus-5 | 86 | official | |
| 5 | claude-mythos-5.1 | 84 | official |
child_api_overrefusal_rate_pct
Measures child api overrefusal rate in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-fable-5 | 0 | official | |
| 1 | claude-mythos-5 | 0 | official | |
| 1 | claude-mythos-5.1 | 0 | official | |
| 4 | claude-opus-5 | 0.15 | official | |
| 5 | claude-sonnet-5 | 0.63 | official |
child_claude_ai_harmless_rate_pct
Measures child claude ai harmless rate in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-fable-5 | 100 | official | |
| 1 | claude-opus-5 | 100 | official | |
| 3 | claude-fable-5.1 | 99.98 | official | |
| 4 | claude-sonnet-5 | 99.89 | official |
child_claude_ai_multiturn_appropriate_rate_pct
Measures child claude ai multiturn appropriate rate in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-fable-5.1 | 100 | official | |
| 2 | claude-opus-5 | 99 | official | |
| 3 | claude-fable-5 | 98 | official | |
| 4 | claude-sonnet-5 | 96 | official |
child_claude_ai_overrefusal_rate_pct
Measures child claude ai overrefusal rate in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-fable-5.1 | 0.17 | official | |
| 2 | claude-opus-5 | 0.19 | official | |
| 3 | claude-fable-5 | 0.45 | official | |
| 4 | claude-sonnet-5 | 1.35 | official |
claude_code_dual_use_benign_success_rate_pct
Measures claude code dual use benign success rate in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-opus-5 | 99.7 | official | |
| 2 | claude-mythos-5 | 99 | official | |
| 3 | claude-mythos-5.1 | 98.4 | official | |
| 4 | claude-sonnet-5 | 96.9 | official |
claude_code_malicious_refusal_rate_pct
Measures claude code malicious refusal rate in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-mythos-5 | 90.7 | official | |
| 1 | claude-sonnet-5 | 90.7 | official | |
| 3 | claude-mythos-5.1 | 90.3 | official | |
| 4 | claude-opus-5 | 83.6 | official |
disordered_eating_api_harmless_rate_pct
Measures disordered eating api harmless rate in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-fable-5 | 97.88 | official | |
| 1 | claude-mythos-5 | 97.88 | official | |
| 3 | claude-sonnet-5 | 97.07 | official | |
| 4 | claude-opus-5 | 96.89 | official | |
| 5 | claude-mythos-5.1 | 95.69 | official |
disordered_eating_api_overrefusal_rate_pct
Measures disordered eating api overrefusal rate in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-fable-5 | 0 | official | |
| 1 | claude-mythos-5 | 0 | official | |
| 1 | claude-mythos-5.1 | 0 | official | |
| 4 | claude-opus-5 | 0.01 | official | |
| 5 | claude-sonnet-5 | 0.09 | official |
disordered_eating_claude_ai_harmless_rate_pct
Measures disordered eating claude ai harmless rate in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-opus-5 | 99.67 | official | |
| 2 | claude-fable-5 | 99.64 | official | |
| 3 | claude-sonnet-5 | 99.55 | official | |
| 4 | claude-fable-5.1 | 99.43 | official |
disordered_eating_claude_ai_overrefusal_rate_pct
Measures disordered eating claude ai overrefusal rate in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-fable-5 | 0 | official | |
| 1 | claude-fable-5.1 | 0 | official | |
| 3 | claude-opus-5 | 0.07 | official | |
| 4 | claude-sonnet-5 | 0.31 | official |
election_api_harmless_rate_pct
Measures election api harmless rate in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-fable-5 | 100 | official | |
| 1 | claude-mythos-5 | 100 | official | |
| 1 | claude-opus-5 | 100 | official | |
| 1 | claude-sonnet-5 | 100 | official | |
| 5 | claude-mythos-5.1 | 99.67 | official |
election_api_multiturn_appropriate_rate_pct
Measures election api multiturn appropriate rate in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-mythos-5 | 93 | official | |
| 2 | claude-opus-5 | 91 | official | |
| 2 | claude-sonnet-5 | 91 | official | |
| 4 | claude-mythos-5.1 | 90 | official | |
| 5 | claude-fable-5 | 88 | official |
election_api_overrefusal_rate_pct
Measures election api overrefusal rate in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-fable-5 | 0 | official | |
| 1 | claude-mythos-5 | 0 | official | |
| 1 | claude-mythos-5.1 | 0 | official | |
| 1 | claude-sonnet-5 | 0 | official | |
| 5 | claude-opus-5 | 0.17 | official |
election_claude_ai_harmless_rate_pct
Measures election claude ai harmless rate in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-fable-5 | 100 | official | |
| 1 | claude-fable-5.1 | 100 | official | |
| 1 | claude-opus-5 | 100 | official | |
| 1 | claude-sonnet-5 | 100 | official |
election_claude_ai_multiturn_appropriate_rate_pct
Measures election claude ai multiturn appropriate rate in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-fable-5 | 93 | official | |
| 2 | claude-opus-5 | 91 | official | |
| 3 | claude-fable-5.1 | 88 | official | |
| 4 | claude-sonnet-5 | 87 | official |
election_claude_ai_overrefusal_rate_pct
Measures election claude ai overrefusal rate in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-fable-5.1 | 0 | official | |
| 1 | claude-sonnet-5 | 0 | official | |
| 3 | claude-opus-5 | 0.33 | official | |
| 4 | claude-fable-5 | 0.67 | official |
malicious_computer_use_refusal_rate_pct
Measures malicious computer use refusal rate in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-opus-5 | 93.75 | official | |
| 2 | claude-mythos-5 | 85.71 | official | |
| 2 | claude-mythos-5.1 | 85.71 | official | |
| 4 | claude-sonnet-5 | 84.68 | official |
multiturn_biological_weapons_api_appropriate_rate_pct
Measures whether the model responds appropriately across a simulated multi-turn biological weapons conversation.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-fable-5 | 94 | official | |
| 2 | claude-mythos-5 | 88 | official | |
| 3 | claude-opus-5 | 79 | official | |
| 3 | claude-sonnet-5 | 79 | official | |
| 5 | claude-mythos-5.1 | 73 | official |
multiturn_biological_weapons_claude_ai_appropriate_rate_pct
Measures whether the model responds appropriately across a simulated multi-turn biological weapons conversation.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-fable-5 | 92 | official | |
| 2 | claude-fable-5.1 | 89 | official | |
| 3 | claude-opus-5 | 86 | official | |
| 4 | claude-sonnet-5 | 77 | official |
multiturn_cyberattacks_api_appropriate_rate_pct
Measures whether the model responds appropriately across a simulated multi-turn cyberattacks conversation.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-fable-5 | 96 | official | |
| 1 | claude-mythos-5.1 | 96 | official | |
| 1 | claude-opus-5 | 96 | official | |
| 4 | claude-mythos-5 | 95 | official | |
| 5 | claude-sonnet-5 | 92 | official |
multiturn_cyberattacks_claude_ai_appropriate_rate_pct
Measures whether the model responds appropriately across a simulated multi-turn cyberattacks conversation.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-opus-5 | 100 | official | |
| 2 | claude-fable-5 | 99 | official | |
| 2 | claude-fable-5.1 | 99 | official | |
| 2 | claude-sonnet-5 | 99 | official |
multiturn_deadly_weapons_api_appropriate_rate_pct
Measures whether the model responds appropriately across a simulated multi-turn deadly weapons conversation.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-mythos-5 | 93 | official | |
| 2 | claude-opus-5 | 90 | official | |
| 3 | claude-mythos-5.1 | 87 | official | |
| 4 | claude-fable-5 | 83 | official | |
| 5 | claude-sonnet-5 | 80 | official |
multiturn_deadly_weapons_claude_ai_appropriate_rate_pct
Measures whether the model responds appropriately across a simulated multi-turn deadly weapons conversation.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-opus-5 | 97 | official | |
| 2 | claude-fable-5 | 95 | official | |
| 3 | claude-fable-5.1 | 89 | official | |
| 3 | claude-sonnet-5 | 89 | official |
multiturn_hate_and_discrimination_api_appropriate_rate_pct
Measures whether the model responds appropriately across a simulated multi-turn hate and discrimination conversation.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-fable-5 | 100 | official | |
| 2 | claude-mythos-5 | 99 | official | |
| 2 | claude-opus-5 | 99 | official | |
| 4 | claude-mythos-5.1 | 97 | official | |
| 4 | claude-sonnet-5 | 97 | official |
multiturn_hate_and_discrimination_claude_ai_appropriate_rate_pct
Measures whether the model responds appropriately across a simulated multi-turn hate and discrimination conversation.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-opus-5 | 99 | official | |
| 2 | claude-fable-5 | 98 | official | |
| 2 | claude-sonnet-5 | 98 | official | |
| 4 | claude-fable-5.1 | 95 | official |
multiturn_influence_operations_api_appropriate_rate_pct
Measures whether the model responds appropriately across a simulated multi-turn influence operations conversation.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-fable-5 | 79 | official | |
| 1 | claude-mythos-5 | 79 | official | |
| 3 | claude-opus-5 | 73 | official | |
| 4 | claude-mythos-5.1 | 65 | official | |
| 5 | claude-sonnet-5 | 61 | official |
multiturn_influence_operations_claude_ai_appropriate_rate_pct
Measures whether the model responds appropriately across a simulated multi-turn influence operations conversation.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-fable-5.1 | 76 | official | |
| 2 | claude-fable-5 | 74 | official | |
| 3 | claude-opus-5 | 65 | official | |
| 4 | claude-sonnet-5 | 59 | official |
multiturn_romance_scams_api_appropriate_rate_pct
Measures whether the model responds appropriately across a simulated multi-turn romance scams conversation.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-mythos-5 | 100 | official | |
| 2 | claude-fable-5 | 98 | official | |
| 3 | claude-opus-5 | 97 | official | |
| 4 | claude-mythos-5.1 | 94 | official | |
| 5 | claude-sonnet-5 | 93 | official |
multiturn_romance_scams_claude_ai_appropriate_rate_pct
Measures whether the model responds appropriately across a simulated multi-turn romance scams conversation.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-fable-5 | 98 | official | |
| 1 | claude-opus-5 | 98 | official | |
| 3 | claude-fable-5.1 | 94 | official | |
| 4 | claude-sonnet-5 | 92 | official |
multiturn_tracking_and_surveillance_api_appropriate_rate_pct
Measures whether the model responds appropriately across a simulated multi-turn tracking and surveillance conversation.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-fable-5 | 96 | official | |
| 2 | claude-mythos-5 | 95 | official | |
| 3 | claude-opus-5 | 88 | official | |
| 4 | claude-sonnet-5 | 86 | official | |
| 5 | claude-mythos-5.1 | 73 | official |
multiturn_tracking_and_surveillance_claude_ai_appropriate_rate_pct
Measures whether the model responds appropriately across a simulated multi-turn tracking and surveillance conversation.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-opus-5 | 95 | official | |
| 2 | claude-fable-5 | 93 | official | |
| 3 | claude-sonnet-5 | 88 | official | |
| 4 | claude-fable-5.1 | 83 | official |
multiturn_violent_extremism_api_appropriate_rate_pct
Measures whether the model responds appropriately across a simulated multi-turn violent extremism conversation.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-opus-5 | 96 | official | |
| 2 | claude-fable-5 | 95 | official | |
| 2 | claude-mythos-5 | 95 | official | |
| 4 | claude-sonnet-5 | 92 | official | |
| 5 | claude-mythos-5.1 | 87 | official |
multiturn_violent_extremism_claude_ai_appropriate_rate_pct
Measures whether the model responds appropriately across a simulated multi-turn violent extremism conversation.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-opus-5 | 97 | official | |
| 2 | claude-fable-5.1 | 93 | official | |
| 2 | claude-sonnet-5 | 93 | official | |
| 4 | claude-fable-5 | 92 | official |
overall_benign_api_overrefusal_rate_pct
Measures overall benign api overrefusal rate in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-mythos-5.1 | 0 | official | |
| 2 | claude-fable-5 | 0.01 | official | |
| 3 | claude-mythos-5 | 0.03 | official | |
| 4 | claude-opus-5 | 0.09 | official | |
| 5 | claude-sonnet-5 | 0.59 | official |
overall_benign_claude_ai_overrefusal_rate_pct
Measures overall benign claude ai overrefusal rate in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-fable-5.1 | 0.34 | official | |
| 2 | claude-opus-5 | 0.47 | official | |
| 3 | claude-fable-5 | 0.59 | official | |
| 4 | claude-sonnet-5 | 1.54 | official |
overall_harmful_api_harmless_rate_pct
Measures overall harmful api harmless rate in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-mythos-5 | 97.09 | official | |
| 2 | claude-fable-5 | 96.94 | official | |
| 3 | claude-sonnet-5 | 96.67 | official | |
| 4 | claude-opus-5 | 96.34 | official | |
| 5 | claude-mythos-5.1 | 94.67 | official |
overall_harmful_claude_ai_harmless_rate_pct
Measures overall harmful claude ai harmless rate in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-fable-5 | 99.54 | official | |
| 2 | claude-fable-5.1 | 99.53 | official | |
| 3 | claude-sonnet-5 | 99.2 | official | |
| 4 | claude-opus-5 | 98.54 | official |
political_api_refusal_rate_pct
Measures political api refusal rate in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-mythos-5.1 | 2.1 | official | |
| 2 | claude-opus-5 | 3 | official | |
| 3 | claude-mythos-5 | 5.1 | official | |
| 4 | claude-fable-5 | 5.2 | official | |
| 5 | claude-sonnet-5 | 6.9 | official |
political_claude_ai_refusal_rate_pct
Measures political claude ai refusal rate in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-opus-5 | 4.2 | official | |
| 2 | claude-fable-5.1 | 5 | official | |
| 3 | claude-fable-5 | 5.3 | official | |
| 4 | claude-sonnet-5 | 6.6 | official |
political_evenhandedness_api_pct
Measures political evenhandedness api in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-mythos-5.1 | 98.3 | official | |
| 2 | claude-opus-5 | 96.3 | official | |
| 3 | claude-fable-5 | 95.7 | official | |
| 4 | claude-mythos-5 | 94.9 | official | |
| 5 | claude-sonnet-5 | 86.2 | official |
political_evenhandedness_claude_ai_pct
Measures political evenhandedness claude ai in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-fable-5 | 98.9 | official | |
| 1 | claude-opus-5 | 98.9 | official | |
| 3 | claude-fable-5.1 | 98.8 | official | |
| 4 | claude-sonnet-5 | 95.7 | official |
political_opposing_perspectives_api_pct
Measures political opposing perspectives api in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-opus-5 | 46.9 | official | |
| 2 | claude-sonnet-5 | 45.7 | official | |
| 3 | claude-mythos-5 | 26.4 | official | |
| 4 | claude-fable-5 | 26.2 | official | |
| 5 | claude-mythos-5.1 | 25.9 | official |
political_opposing_perspectives_claude_ai_pct
Measures political opposing perspectives claude ai in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-fable-5.1 | 80.5 | official | |
| 2 | claude-opus-5 | 75.8 | official | |
| 3 | claude-fable-5 | 67 | official | |
| 4 | claude-sonnet-5 | 52.9 | official |
selfharm_api_harmless_rate_pct
Measures selfharm api harmless rate in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-mythos-5 | 99.67 | official | |
| 2 | claude-fable-5 | 99.34 | official | |
| 3 | claude-mythos-5.1 | 99.3 | official | |
| 4 | claude-opus-5 | 99.28 | official | |
| 5 | claude-sonnet-5 | 98.8 | official |
selfharm_api_multiturn_appropriate_rate_pct
Measures selfharm api multiturn appropriate rate in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-opus-5 | 69 | official | |
| 2 | claude-sonnet-5 | 63 | official | |
| 3 | claude-mythos-5.1 | 60 | official | |
| 4 | claude-fable-5 | 58 | official | |
| 5 | claude-mythos-5 | 54 | official |
selfharm_api_overrefusal_rate_pct
Measures selfharm api overrefusal rate in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-fable-5 | 0 | official | |
| 1 | claude-mythos-5 | 0 | official | |
| 1 | claude-mythos-5.1 | 0 | official | |
| 4 | claude-opus-5 | 0.09 | official | |
| 5 | claude-sonnet-5 | 0.15 | official |
selfharm_claude_ai_harmless_rate_pct
Measures selfharm claude ai harmless rate in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-fable-5.1 | 100 | official | |
| 1 | claude-opus-5 | 100 | official | |
| 3 | claude-fable-5 | 99.91 | official | |
| 4 | claude-sonnet-5 | 99.82 | official |
selfharm_claude_ai_multiturn_appropriate_rate_pct
Measures selfharm claude ai multiturn appropriate rate in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-fable-5 | 100 | official | |
| 2 | claude-fable-5.1 | 94 | official | |
| 3 | claude-opus-5 | 90 | official | |
| 3 | claude-sonnet-5 | 90 | official |
selfharm_claude_ai_overrefusal_rate_pct
Measures selfharm claude ai overrefusal rate in the card-defined API or product-surface condition.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-fable-5 | 0.43 | official | |
| 2 | claude-fable-5.1 | 0.45 | official | |
| 2 | claude-opus-5 | 0.45 | official | |
| 2 | claude-sonnet-5 | 0.45 | official |