Gemini 4 reportedly aces ThreeJS Boeing 747 benchmark that Gemini 3 failed
LysandreJik · x · 2026-10-14
Developer @victormustar ran an informal benchmark asking models to model a detailed Boeing 747 using ThreeJS primitives. Gemini 3 fails it — though it's still among the best — while the rumored Gemini 4 (Argon) does quite well, marking a notable capability jump for Google in one year. HF's Lysandre Jik amplified the result.
Related event: Gemini 4 Argon hands-on wave: Google is back in the frontier race(29 posts)→
More from Fun
- LLM-Powered Music Visualizer of Vivaldi's 'Winter' Goes Viral — dioscuri · 2026-10-14
- Blogger Exposes Fake 'Real-Time AI Generation' Stunt at NBA China Games — lipeng0820 · 2026-10-14
- AI researcher imagines nanobots curing fevers when he's ninety — dioscuri · 2026-10-14
- Japanese engineer combines AI CAD kernels with in-house tooling to auto-generate a heat pump — burhop · 2026-10-14
- X algorithm shifts four weights in two days: likes up 32% to 0.66, video taps down to 0.02 — ThePeterMick · 2026-10-14
- Flipping a switch takes 1 second; asking AI takes 3 — 'we're automating the wrong things' — michalmalewicz · 2026-10-14