Benchmark	Category	Top Score	Info	Self Reported	Source
AIME 2025	math	1	Avg@5, 64k Thinking, With Tools	Yes	Source
ARC-AGI-1	-	0.80	64k Thinking	No	Source
ARC-AGI-2	-	0.38	64k Thinking	No	Source
GPQA Diamond	general-knowledge	0.87	Avg@5, 64k Thinking	Yes	Source
Humanity's Last Exam	-	0.43	With Search	Yes	Source
MMMLU	-	0.91	Avg@5, 64k Thinking	Yes	Source
MMMU	-	0.81	Avg@5, 64k Thinking	Yes	Source
OSWorld	-	0.66	Pass@1; Avg@5, 64k Thinking	Yes	Source
SWE Bench Multilingual	code	0.76	Avg@5	Yes	Source
SWE Bench Pro	-	0.52	Avg@5	Yes	Source
SWE-Bench	code	0.81	Avg@5, 64k Thinking	Yes	Source
Tau 2 Airline	-	0.88	Corrected	Yes	Source
Tau 2 Retail	-	0.89	Avg@5, 64k Thinking	Yes	Source
Tau 2 Telecom	-	0.98	Avg@5, 64k Thinking	Yes	Source
Terminal Bench 2.0	-	0.59	Avg@5, 128k Thinking	Yes	Source
Vending Bench 2	-	4967	8k Thinking	No	Source

Use the selector to switch benchmarks and see how this model stacks up against its closest competitors.

Compare this model with the leading peers for the selected benchmark.

Benchmark

0.81

Rank #1/21

21 models

Showing 11 models around the selected model (out of 21 total).

Preparing your page

Claude Opus 4.5