claudeVIDEO
Blender benchmark: GPT-6 Astra scores 0.686 vs Opus 5.5's 0.642
Across 20 Blender tasks, GPT-6 Astra averaged 0.686 reward versus 0.642 for Claude Opus 5.5, with Astra scoring 0.90+ on 7 tasks and Opus on 3; on a tightrope task Opus kept the cowboy's feet planted and scored 0.400.
Refresh
@refresh_devArchived 2026.10.01Originally posted 2026.10.01
Refresh @refresh_devOriginally posted 2026.10.01
GPT-6 Astra averaged 0.686 reward across 20 Blender tasks, Claude Opus 5.5 0.642. Astra hit 0.90+ on 7/20, Opus on 3/20. Where they fall short: on this tightrope task, Opus 5.5 kept the cowboy's feet planted instead of stepping and scored 0.400.
x.com/refresh_dev/status/2105376464648773946Archived 2026.10.01 10:16