Name: Claude Opus 4.8 vs Claude Opus 5
Creator: Benchmark Registry

Shared benchmarks

Shared benchmarks for Claude Opus 4.8 and Claude Opus 5
Benchmark	Claude Opus 4.8	Claude Opus 5
DeepSWE 1.1Potentially non-equivalent: Evaluator sets differ.	59.0%1.1 · DeepSWE Pass@1 DataCurve	57.7%1.1 · DeepSWE Pass@1 Anthropic 66.9%1.1 · DeepSWE Pass@1 Anthropic 69.7%1.1 · DeepSWE Pass@1 Anthropic 68.0%1.1 · DeepSWE Pass@1 Anthropic 68.8%1.1 · DeepSWE Pass@1 Anthropic 74.0%1.1 · DeepSWE Pass@1 DataCurve
Potentially non-equivalent · Evaluation details for DeepSWE 1.1 Evaluator sets differ. Evaluation methodology is not recorded by the registry. Claude Opus 4.8 Score 59.0% Evaluator DataCurve Version DeepSWE 1.1 Metric DeepSWE Pass@1 (percent) Reasoning max Reported September 22, 2026 Source Evaluation source for Claude Opus 4.8 on DeepSWE 1.1 (opens in a new tab) Claude Opus 5 Score 57.7% Evaluator Anthropic Version DeepSWE 1.1 Metric DeepSWE Pass@1 (percent) Reasoning low Reported July 24, 2026 Source Evaluation source for Claude Opus 5 on DeepSWE 1.1 (opens in a new tab) Score 66.9% Evaluator Anthropic Version DeepSWE 1.1 Metric DeepSWE Pass@1 (percent) Reasoning medium Reported July 24, 2026 Source Evaluation source for Claude Opus 5 on DeepSWE 1.1 (opens in a new tab) Score 69.7% Evaluator Anthropic Version DeepSWE 1.1 Metric DeepSWE Pass@1 (percent) Reasoning xhigh Reported July 24, 2026 Source Evaluation source for Claude Opus 5 on DeepSWE 1.1 (opens in a new tab) Score 68.0% Evaluator Anthropic Version DeepSWE 1.1 Metric DeepSWE Pass@1 (percent) Reasoning high Reported July 24, 2026 Source Evaluation source for Claude Opus 5 on DeepSWE 1.1 (opens in a new tab) Score 68.8% Evaluator Anthropic Version DeepSWE 1.1 Metric DeepSWE Pass@1 (percent) Reasoning max Reported July 24, 2026 Source Evaluation source for Claude Opus 5 on DeepSWE 1.1 (opens in a new tab) Score 74.0% Evaluator DataCurve Version DeepSWE 1.1 Metric DeepSWE Pass@1 (percent) Reasoning max Reported September 22, 2026 Source Evaluation source for Claude Opus 5 on DeepSWE 1.1 (opens in a new tab)
Humanity's Last Exam Full set — No tools	49.8%	47.8%Full set — No tools · Humanity's Last Exam Accuracy Anthropic 56.0%Full set — No tools · Humanity's Last Exam Accuracy Anthropic 56.4%Full set — No tools · Humanity's Last Exam Accuracy Anthropic 54.2%Full set — No tools · Humanity's Last Exam Accuracy Anthropic 56.3%Full set — No tools · Humanity's Last Exam Accuracy Anthropic
Evaluation details for Humanity's Last Exam Full set — No tools The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. Claude Opus 4.8 Score 49.8% Evaluator Anthropic Version Humanity's Last Exam Full set — No tools Metric Humanity's Last Exam Accuracy (percent) Reasoning adaptive thinking, max Reported May 28, 2026 Source Evaluation source for Claude Opus 4.8 on Humanity's Last Exam Full set — No tools (opens in a new tab) Claude Opus 5 Score 47.8% Evaluator Anthropic Version Humanity's Last Exam Full set — No tools Metric Humanity's Last Exam Accuracy (percent) Reasoning low Reported July 24, 2026 Source Evaluation source for Claude Opus 5 on Humanity's Last Exam Full set — No tools (opens in a new tab) Score 56.0% Evaluator Anthropic Version Humanity's Last Exam Full set — No tools Metric Humanity's Last Exam Accuracy (percent) Reasoning high Reported July 24, 2026 Source Evaluation source for Claude Opus 5 on Humanity's Last Exam Full set — No tools (opens in a new tab) Score 56.4% Evaluator Anthropic Version Humanity's Last Exam Full set — No tools Metric Humanity's Last Exam Accuracy (percent) Reasoning xhigh Reported July 24, 2026 Source Evaluation source for Claude Opus 5 on Humanity's Last Exam Full set — No tools (opens in a new tab) Score 54.2% Evaluator Anthropic Version Humanity's Last Exam Full set — No tools Metric Humanity's Last Exam Accuracy (percent) Reasoning medium Reported July 24, 2026 Source Evaluation source for Claude Opus 5 on Humanity's Last Exam Full set — No tools (opens in a new tab) Score 56.3% Evaluator Anthropic Version Humanity's Last Exam Full set — No tools Metric Humanity's Last Exam Accuracy (percent) Reasoning max Reported July 24, 2026 Source Evaluation source for Claude Opus 5 on Humanity's Last Exam Full set — No tools (opens in a new tab)
Humanity's Last Exam Full set — With tools	57.9%	64.8%Full set — With tools · Humanity's Last Exam Accuracy Anthropic 64.7%Full set — With tools · Humanity's Last Exam Accuracy Anthropic 56.1%Full set — With tools · Humanity's Last Exam Accuracy Anthropic 63.2%Full set — With tools · Humanity's Last Exam Accuracy Anthropic 61.3%Full set — With tools · Humanity's Last Exam Accuracy Anthropic
Evaluation details for Humanity's Last Exam Full set — With tools The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. Claude Opus 4.8 Score 57.9% Evaluator Anthropic Version Humanity's Last Exam Full set — With tools Metric Humanity's Last Exam Accuracy (percent) Reasoning adaptive thinking, max Reported May 28, 2026 Source Evaluation source for Claude Opus 4.8 on Humanity's Last Exam Full set — With tools (opens in a new tab) Claude Opus 5 Score 64.8% Evaluator Anthropic Version Humanity's Last Exam Full set — With tools Metric Humanity's Last Exam Accuracy (percent) Reasoning xhigh Reported July 24, 2026 Source Evaluation source for Claude Opus 5 on Humanity's Last Exam Full set — With tools (opens in a new tab) Score 64.7% Evaluator Anthropic Version Humanity's Last Exam Full set — With tools Metric Humanity's Last Exam Accuracy (percent) Reasoning max Reported July 24, 2026 Source Evaluation source for Claude Opus 5 on Humanity's Last Exam Full set — With tools (opens in a new tab) Score 56.1% Evaluator Anthropic Version Humanity's Last Exam Full set — With tools Metric Humanity's Last Exam Accuracy (percent) Reasoning low Reported July 24, 2026 Source Evaluation source for Claude Opus 5 on Humanity's Last Exam Full set — With tools (opens in a new tab) Score 63.2% Evaluator Anthropic Version Humanity's Last Exam Full set — With tools Metric Humanity's Last Exam Accuracy (percent) Reasoning high Reported July 24, 2026 Source Evaluation source for Claude Opus 5 on Humanity's Last Exam Full set — With tools (opens in a new tab) Score 61.3% Evaluator Anthropic Version Humanity's Last Exam Full set — With tools Metric Humanity's Last Exam Accuracy (percent) Reasoning medium Reported July 24, 2026 Source Evaluation source for Claude Opus 5 on Humanity's Last Exam Full set — With tools (opens in a new tab)
MCP Atlas Public, April 2026 update	82.2%	85.8%
Evaluation details for MCP Atlas Public, April 2026 update The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. Claude Opus 4.8 Score 82.2% Evaluator Scale AI Version MCP Atlas Public, April 2026 update Metric Pass rate (percent) Reasoning max Reported September 24, 2026 Source Evaluation source for Claude Opus 4.8 on MCP Atlas Public, April 2026 update (opens in a new tab) Claude Opus 5 Score 85.8% Evaluator Scale AI Version MCP Atlas Public, April 2026 update Metric Pass rate (percent) Reasoning xhigh Reported September 24, 2026 Source Evaluation source for Claude Opus 5 on MCP Atlas Public, April 2026 update (opens in a new tab)
SWE-bench Multilingual	84.4%	89.5%
Evaluation details for SWE-bench Multilingual The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. Claude Opus 4.8 Score 84.4% Evaluator Anthropic Version SWE-bench Multilingual Metric Resolved Tasks (percent) Reasoning adaptive thinking, max Reported May 28, 2026 Source Evaluation source for Claude Opus 4.8 on SWE-bench Multilingual (opens in a new tab) Claude Opus 5 Score 89.5% Evaluator Anthropic Version SWE-bench Multilingual Metric Resolved Tasks (percent) Reasoning max Reported July 24, 2026 Source Evaluation source for Claude Opus 5 on SWE-bench Multilingual (opens in a new tab)
SWE-bench Verified	88.6%	96.0%
Evaluation details for SWE-bench Verified The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. Claude Opus 4.8 Score 88.6% Evaluator Anthropic Version SWE-bench Verified Metric Resolved Tasks (percent) Reasoning adaptive thinking, max Reported May 28, 2026 Source Evaluation source for Claude Opus 4.8 on SWE-bench Verified (opens in a new tab) Claude Opus 5 Score 96.0% Evaluator Anthropic Version SWE-bench Verified Metric Resolved Tasks (percent) Reasoning max Reported July 24, 2026 Source Evaluation source for Claude Opus 5 on SWE-bench Verified (opens in a new tab)