OSWorld 2.0

AI benchmark from Anthropic System Card

Full listing name: OSWorld 2.0 (Anthropic evaluation)

Anthropic's first-attempt evaluation of Claude Opus 5 on OSWorld 2.0 tasks. This variant preserves the system card's reported metric separately from the benchmark paper's binary-completion setup.

Publisher
Anthropic System Card
Domains
Computer Use
Data source
benchmarklist
Catalogued
Website
https://benchmarklist.com/benchmarks/osworld_2_0_anthropic/

Catalogue entry from the RL Engineering daily scrape of public sources. Data reflects the 2026-08-30 snapshot.