Same prompt, nine outputs: watching Anthropic models evolve from Opus 4.5 to Opus 5.5 (Max)
arena · x · 2026-11-19
LMArena highlighted petergostev's experiment generating nine outputs with the identical prompt across Anthropic models from Opus 4.5 to Opus 5.5 (Max), a side-by-side look at generational progress. The prompts are open-sourced in the 3d-prompt-collection GitHub repo (414 stars), and anyone can rerun the comparison on Arena.
More from Models
- Pro subscriber complains coding tool quotas allow only 3 questions per 5 hours — teortaxesTex · 2026-11-19
- Users report 'Context window full' locks forcing new ChatGPT chats, paid users included — ombre-ai · 2026-11-19
- Hands-on coding test: Tencent's Hy4 Flash beats DeepSeek by 12x on same task — HeyToha · 2026-11-19
- Same weights, 3x score: Gemini 4 Neon jumps 12.8%→39% on SWE-sweep by swapping the harness — Delicious_Character6 · 2026-11-19
- Testing gemini-4-neon with a DIY PowerShell agent harness: "benchmarks are for tourists" — dfinke · 2026-11-19
- Lecture on Gemini 4 post-training boils down to one thing: high-quality data — mariofilhoml · 2026-11-19