Claude Sonnet 4 vs Claude Sonnet 4.5
13 shared benchmark contexts with reported scores and primary source links.
Claude Sonnet 4Claude Sonnet 4.5
Change models in the comparison toolShared benchmarks
13| Benchmark | Claude Sonnet 4 | Claude Sonnet 4.5 |
|---|---|---|
|
| |
Evaluation details for GPQA Diamond — Epoch AI harnessThe recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. | ||
|
| |
Evaluation details for GPQA Diamond — Vals AI harnessThe recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. | ||
|
| |
Evaluation details for Humanity's Last Exam Full set — Scale AI evaluationThe recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. Claude Sonnet 4 | ||
|
| |
Evaluation details for Humanity's Last Exam Text only — Scale AI evaluationThe recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. Claude Sonnet 4 | ||
|
| |
Evaluation details for MMLU-Pro Original — Vals AI harnessThe recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. | ||
|
| |
Evaluation details for MMMU-Pro Vals AI harnessThe recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. | ||
|
| |
Evaluation details for OSWorld Verified — 100 stepsThe recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. Claude Sonnet 4 Claude Sonnet 4.5 | ||
|
| |
Evaluation details for OSWorld Verified — 15 stepsThe recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. Claude Sonnet 4 Claude Sonnet 4.5 | ||
|
| |
Evaluation details for OSWorld Verified — 50 stepsThe recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. Claude Sonnet 4 Claude Sonnet 4.5 | ||
|
| |
Evaluation details for SWE-bench VerifiedThe recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. Claude Sonnet 4 Claude Sonnet 4.5 | ||
|
| |
Evaluation details for SWE-bench Verified — mini-swe-agentThe recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. | ||
|
| |
Evaluation details for SWE-bench Pro Public — Scale AI SWE-AgentThe recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. | ||
|
| |
Evaluation details for Toolathlon OriginalThe recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. Claude Sonnet 4 Claude Sonnet 4.5 | ||