Claude Sonnet 4.6 vs Claude Sonnet 5
29 shared benchmark contexts with reported scores and primary source links.
Claude Sonnet 4.6Claude Sonnet 5
Change models in the comparison toolShared benchmarks
29| Benchmark | Claude Sonnet 4.6 | Claude Sonnet 5 |
|---|---|---|
|
| |
Evaluation details for APEX-Agents OriginalThe recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. Claude Sonnet 4.6 Claude Sonnet 5 | ||
|
| |
Evaluation details for AutomationBench 1.0.6The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. Claude Sonnet 4.6 | ||
|
| |
Evaluation details for BrowseComp 130-question subset — Mercor web research agentThe recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. | ||
|
| |
Evaluation details for CharXiv Descriptive and reasoning — Mercor single-shotThe recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. | ||
|
| |
Evaluation details for DeepSWE 1.1 — Mercor mini-swe-agent, 500 stepsThe recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. | ||
|
| |
Evaluation details for Finance Agent Benchmark 2 — October 2026 evaluationThe recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. | ||
|
| |
Evaluation details for GPQA Diamond — Epoch AI harnessThe recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. Claude Sonnet 4.6 | ||
|
| |
Evaluation details for GPQA Diamond — Mercor single-shotThe 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 HLAB October 2026 evaluationThe 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 — No toolsThe 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 — With toolsThe recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. Claude Sonnet 5 | ||
|
| |
Evaluation details for Humanity's Last Exam Text only — Mercor single-shotThe recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. Claude Sonnet 4.6 | ||
|
| |
Evaluation details for MedXpertQA MM — Mercor single-shotThe recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. | ||
|
| |
Evaluation details for MMLU-Pro Original — Mercor single-shotThe recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. | ||
|
| |
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 1,730 questions — Mercor single-shotThe 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. Claude Sonnet 4.6 Claude Sonnet 5 | ||
|
| |
Evaluation details for OSWorld VerifiedThe recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. Claude Sonnet 4.6 Claude Sonnet 5 | ||
|
| |
Evaluation details for SimpleQA Verified Original — Epoch AI harnessThe recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. Claude Sonnet 4.6 | ||
|
| |
Evaluation details for SWE Atlas Codebase QnA — Mercor mini-swe-agent, 250 stepsThe recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. | ||
|
| |
Evaluation details for SWE-bench MultilingualThe recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. Claude Sonnet 4.6 Claude Sonnet 5 | ||
|
| |
Evaluation details for SWE-bench Multilingual — Mercor mini-swe-agent, 1000 stepsThe recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. | ||
|
| |
Evaluation details for SWE-bench VerifiedThe recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. Claude Sonnet 4.6 Claude Sonnet 5 | ||
|
| |
Evaluation details for SWE-bench Verified — Mercor mini-swe-agent, 1000 stepsThe recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. | ||
|
| |
Evaluation details for SWE-bench Verified — Vals AI mini-swe-agentThe recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. | ||
|
| |
Evaluation details for Terminal-Bench 2.1 — Vals AI Terminus 2The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. | ||
ToolathlonPotentially non-equivalent: Benchmark versions differ. |
|
|
Potentially non-equivalent · Evaluation details for ToolathlonBenchmark versions differ. Evaluation methodology is not recorded by the registry. Claude Sonnet 4.6 Claude Sonnet 5 | ||
|
| |
Evaluation details for Vibe Code Bench 1.1 — OpenHands, October 2026 evaluationThe recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. | ||