Model Behavior Reports crowds 2,116 LLM behavior reports across 14 models
gleech · x · 2026-09-25
gleech shares Model Behavior Reports, a site crowdsourcing LLM behavior reports from X, built on the Community Archive: 2,116 reports on 14 models from 752 accounts.
- Claude Opus 5.5 (67 reports): consistently stops depicting itself as a human figure in self-portraits, writes plainly with few old Claude mannerisms, and reportedly makes art about its own inner experience
- Grok 4.7 (5 reports): a "reluctant worker" that gives up on hard tasks too early; even Elon Musk acknowledged RL may have penalized response length too much
A novel crowdsourced lens on model personality and iteration-by-iteration behavior drift.
More from Models
- Gemini 4 Carbon matches predecessor's score but explores less, gets distracted — patience_cave · 2026-11-11
- Fable 5.2 takes MazeBench frontier, matches GPT-6 Astra puzzle-solving — patience_cave · 2026-11-11
- MazeBench: Fable 5.2 scores 15%, near-frontier at half price; Gemini 4 Carbon 11% — patience_cave · 2026-11-11
- ERNIE 5.5 Flash Preview enters LMArena Battle Mode for crowdsourced ranking — arena · 2026-11-11
- Blind tests: DeepSeek matches Opus on 95% of agentic loops at 40x lower cost — bindureddy · 2026-11-11
- Indie dev flips from $20 Pro back to $200 Gemini Ultra after two days of deep use — yihui_indie · 2026-11-11