OnlyModels.gg stages LLM fight sims; all ten fighters tie at 80 HP
Dave_Maynor · x · 2026-11-08
OnlyModels.gg runs a daily "Model Fight" where LLMs battle as virtual fight cards — actually testing their eval execution and escape-claim abilities. Day 3's card of ten Qwen-derived models ended in a ten-way 80 HP tie, with most fighters backfiring by returning no parseable evals; even some verifiers failed. A post-brawl APEX escape-claim benchmark was also run. Absurd format, but essentially a model-behavior eval toy.
Related event: OnlyModels.gg Pits LLMs in Nightly Fight Club as Eval Test(2 posts)→
More from Fun
- User Vows Never to Like an Influencer's Close-Friends Thirst Traps — EddyVGG · 2026-11-08
- Kalshi jokes Dario Amodei testified he doesn't know his own home address — CtrlAltDwayne · 2026-11-08
- Pedro Domingos: Anthropic builds anthropomorphic AI — and that's what makes it dangerous — pmddomingos · 2026-11-08
- yacineMTB memes on overengineering: why many code when few code solve problem? — yacineMTB · 2026-11-08
- Emergent UI Downside: Colleagues Now Expect You to Read Every AI-Generated Interface — sloppenheimer · 2026-11-08
- Has anyone named their baby Opus yet? Or Codex? — SIGKITTEN · 2026-11-08