Animation Bench: GPT-6.1 Sol tops coding agents on web animation reconstruction at $0.46/task
himanshustwts · x · 2026-10-02
Physera released Animation Bench v1.0, testing frontier coding agents on reconstructing behavior (not just looks) of 48 real web animations from 32 commercial sites, 240 reconstructions total. Key results:
- Leaderboard: GPT-6.1 Sol (0.606, Motion 0.512), GPT-6 Astra (0.594), Claude Fable 5.1 (0.548), GPT-6 Sol (0.516), Claude Opus 5.5 (0.507)
- GPT-6.1 Sol won 14 of 48 tasks at $0.46/task, far cheaper than GPT-6 Astra's $3.04; GPT-6 Sol costs $0.45
- Caveat: Sol 6.1 was added Oct 1 under the same tasks/agent/scorer, but recording conditions alone can shift the 48-task mean by 0.01, so its lead over Astra is within uncertainty
- Scores are 0-1 reproduction quality, not success rates; the site offers frame-locked reference vs. model overlays
Related event: GPT-6.1 Sol Tops Animation Bench for Web Animation Reconstruction(2 posts)→
More from coding & agent
- GPT-6.1 Astra builds Gaussian Splats via Blender MCP, viewed in SuperSplat — willeastcott · 2026-10-02
- Hoshikage Pulse: an Opus 5.5-built rhythm game with 128 songs, playable in browser — TAbrodi · 2026-10-02
- Conductor Mobile launches: run and control a team of cloud coding agents from iPhone — charlieholtz · 2026-10-02
- Do AI agents actually save time, or do you spend more time fixing their mistakes? — MantisReka · 2026-10-02
- Codex users hit imagegen skill installed but image_gen tool unavailable — Reasonable-Donut-866 · 2026-10-02
- Open-source Hermes ChatGPT Extension embeds Hermes agents inside Codex — intellectronica · 2026-10-02