Slinky-down-escalator test: Claude Opus 5.5 beats Fable 5.2 at realistic perpetual motion
shniydder · reddit · 2026-11-11
Reddit user shniydder reran their "rainbow slinky walking down an up escalator" benchmark on Claude Fable 5.2 (xhigh) vs Claude Opus 5.5 (xhigh):
- Task: a single HTML file with a physically realistic (not scripted) perpetually walking slinky, plus start/stop/reset buttons
- Screenshots: Opus 5.5 built in 1h14m with a more coherent slinky; Fable 5.2 finished in 1h04m but the geometry visibly collapses
- Verdict: Opus 5.5 still better at most things you care about; all other models tested did worse
- Author argues informal "real frontier" demos beat TMB (Trust Me Bro) benchmarks, and jokes Anthropic could burn these weights into chips if they just removed the ugly text watermarking
More from Models
- gnukeith details roadmap: alignment model, main model, decision model, then sub-9B specialists — gnukeith · 2026-11-11
- Indie dev gnukeith teases free self-trained LLM landing mid-January with a companion alignment model — gnukeith · 2026-11-11
- Yuanbao-Reranker: 0.6B/4B/8B Qwen3-based rerankers trained on synthetic preferences, open-sourced — _reachsumit · 2026-11-11
- Indie trainer gnukeith bets on sub-9B models: a 2B model can likely do 75% of your tasks — gnukeith · 2026-11-11
- Leaked chart claims Anthropic's internal Model 4 hit 178.76 AECI — haider1 · 2026-11-11
- Dev trains small CoreML decision model, beats his baselines by a wide margin — cyrilzakka · 2026-11-11