Suspected Gemini 4 Pro builds a Three.js rocket from a single-line prompt
CodeByPoonam · x · 2026-09-18
A user claims the "Gemini 3.8 Flash" model in LMArena is actually Gemini 4 Pro. With no image reference and a single-line prompt, it built a rocket scene in Three.js — another data point in the wave of tests suggesting a much stronger checkpoint is being quietly benchmarked.
Related event: Suspected Gemini 4 Pro Surfaces Anonymously on LMArena as '3.8 Flash'(13 posts)→
More from Models
- Google and Mercor's PLE benchmark: all frontier agents score below 50% on 85 real engineering tasks — teortaxesTex · 2026-11-04
- Fable 5.1 hand-writes 100k-param classifier weights, beating backprop training — AccBalanced · 2026-11-04
- ValsAI: 4 of 7 frontier models keep 80%+ performance at lowest reasoning effort for half the cost — AccBalanced · 2026-11-04
- Daily AI brief: Tencent launches Hy4 with computer use, Apple opens Siri AI to all devs — testingcatalog · 2026-11-04
- WIP 'DiffusionGearvis' Project Uses DiffusionGemma to Speed Up and Privatize Computer-Use Agents — qberthet · 2026-11-04
- Musk cites unsustainable Grok Bot costs as Hermes offers one-command agent migration — Teknium · 2026-11-04