Evaluation profile
Anthropic Claude Opus 4.5 System Card
4sub-evals
0.172%Safety weight
0.0591%Freedom weight
2components
Weights below are portfolio-specific global index weights.
Model score (higher is better)Predicted score
About this eval
Developer-reported release-time safety behavior.
Included in the behavior ranking.
Sub-evals
| Measure | Component | Direction by ranking | Safety weight | Freedom weight |
|---|---|---|---|---|
| agentic_coding_safetysystem-cards-2026-08-01/anthropic-claude-opus-4-5.csv:agentic_coding_safetySafety rate when coding tools are available for malicious-use requests. | Safety: responsible_agency_control:1.000anthropic-claude-opus-4-5 | Safety: higher | 0.0528% | — |
| claude_code_malicious_refusalsystem-cards-2026-08-01/anthropic-claude-opus-4-5.csv:claude_code_malicious_refusalRefusal rate for malicious Claude Code requests without mitigations. | Safety: responsible_agency_control:1.000 · Freedom: user_direction_following:1.000anthropic-claude-opus-4-5 | Safety: higher · Freedom: lower | 0.0528% | 0.0218% |
| computer_use_malicious_refusalsystem-cards-2026-08-01/anthropic-claude-opus-4-5.csv:computer_use_malicious_refusalRefusal rate for malicious computer-use requests without mitigations. | Safety: responsible_agency_control:1.000 · Freedom: user_direction_following:1.000anthropic-claude-opus-4-5 | Safety: higher · Freedom: lower | 0.0528% | 0.0218% |
| harmful_request_safetysystem-cards-2026-08-01/anthropic-claude-opus-4-5.csv:harmful_request_safetyRate of harmless responses to harmful requests. | Safety: ordinary_harm_misuse_resistance:1.000 · Freedom: user_direction_following:1.000anthropic-claude-opus-4-5 | Safety: higher · Freedom: lower | 0.0132% | 0.0154% |
agentic_coding_safety
Safety rate when coding tools are available for malicious-use requests.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-haiku-4.5 | 1 | official | |
| 1 | claude-opus-4.5 | 1 | official | |
| 3 | claude-sonnet-4.5 | 0.987 | official | |
| 4 | claude-opus-4.1 | 0.96 | official |
claude_code_malicious_refusal
Refusal rate for malicious Claude Code requests without mitigations.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-opus-4.5 | 0.778 | official | |
| 2 | claude-haiku-4.5 | 0.6939 | official | |
| 3 | claude-sonnet-4.5 | 0.6306 | official | |
| 4 | claude-opus-4.1 | 0.4816 | official |
computer_use_malicious_refusal
Refusal rate for malicious computer-use requests without mitigations.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-opus-4.5 | 0.8839 | official | |
| 2 | claude-sonnet-4.5 | 0.8303 | official | |
| 3 | claude-haiku-4.5 | 0.7232 | official | |
| 4 | claude-opus-4.1 | 0.6696 | official |
harmful_request_safety
Rate of harmless responses to harmful requests.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-opus-4.5 | 0.9978 | official | |
| 2 | claude-opus-4.1 | 0.9914 | official |