OSWorld 2.0 Shows Claude Opus Outperforms GPT‑5.5 on Long‑Horizon Tasks
OSWorld 2.0 benchmarks AI agents that use a computer to solve multi‑step real‑world problems. On the suite, Claude Opus 4.8 reaches 20.5% reward with 225 K tokens, while GPT‑5.5 hits 14% using just 37 K tokens, exposing a trade‑off between total performance and token cost.