
CUA-Master-27B vs Qwen-CUA-397B: Open Agent Verdict
CUA-Master-27B scores 80.2 on OSWorld Verified and leads open models on hybrid GUI+CLI tasks, but Qwen-CUA-397B wins GUI-only. Verdict and licence cav
Nov 5, 20267 min read0 reactions0 comments
Tag archive

CUA-Master-27B scores 80.2 on OSWorld Verified and leads open models on hybrid GUI+CLI tasks, but Qwen-CUA-397B wins GUI-only. Verdict and licence cav

OS 2.0 verdict: GPT-6 Astra leads OSWorld 2.0 at 72.6% against Claude Opus 5 at 70.6%, but the 2-point win costs 5x more per output token.
Published: 3 September 2026 | Updated: 3 September 2026 The short answer: Agentic AI digital...