chiprook
← AI
AISeptember 29, 2026, 10:43

GPT-6 Astra beats Claude Opus 5 on OSWorld 2.0 benchmark

On the OSWorld 2.0 long-horizon computer-use benchmark, GPT-6 Astra scored 72.6% binary completion versus 70.6% for Claude Opus 5, according to the Snorkel AI leaderboard. Astra costs $10 per million input tokens and $50 per million output tokens, and is OpenAI's first model to hit the Critical cybersecurity threshold, so enterprise access is off by default.

#OpenAI#Anthropic#GPT-6#Claude
Read next
AI

Opus 5.5 Beats GPT-6 Astra in 8 of 12 Real-World Trials

AI

Claude Opus 5.5 vs GPT-6 Astra vs GPT-6 Sol: effort setting matters more than model

AI

Real-SWE: Fable 5.1 beats GPT-6 Astra and Gemini 3.8 Flash on coding

AI

Leak: Gemini 4 Pro Reportedly Beats Opus 5.5 and GPT6 Astra