Benchmark

Anthropic — reported benchmarks

Anthropic · source page ↗ · last checked Nov 9, 2026, 12:04 PM

Reported benchmarks · Claude Opus 5.5

captured Nov 9, 2026, 12:04 PM
BenchmarkScore
Terminal-Bench 4.066.4% · xhigh effort; standard error ±2.6 pts
FrontierCode v1.1 (Main)54.4%
CursorBench 4.057.8%
SWE-bench Pro89.9%
SWE-bench Multilingual93.9%
SWE-bench Multimodal61.4%
GDPval-AA v2.11846 Elo · knowledge work benchmark
Humanity's Last Exam67.7% · with tools
Terminal-Bench-Science 0.158.7%
OSWorld 2.081.8% · partial credit
ArXivMath91.2% · without tools
Chartography89.0% · with tools

Vendor-reported via automated web search — not independently verified. See the cited matrix on /models.

In the news · Anthropic

Importance-filtered press coverage (Google News) mentioning Anthropic. Headlines link to the original; verify before acting.

Change history