SWE-Atlas-QnA grading is revised, now SWE-Atlas-QnA-AA. Scale AI's 124 repository Q&A tasks are unchanged. We reworded 127 of the 1,306 rubric criteria for more consistent verdicts and moved to GPT-6 Luna (xhigh) as the judge, with five votes per criterion.
Muse Code with Muse Wonder Full (max) leads at 68.8%, followed by Claude Code with Claude Fable 5.1 (max with fallback) at 66.4%, Claude Code with Claude Opus 5.5 (max) at 65.9% and Claude Code with Claude Sonnet 5.5 (max) at 64.2%. Across the 21 configurations evaluated on both Index versions, the mean score falls 1.7 points, with 14 of 21 moving down.