Arena Reportedly Testing Gemini 4 Pro Under gemini-3.8-flash Codename
CodeByPoonam · x · 2026-09-27
A user reports that a model routed to them on LMArena under the name gemini-3.8-flash appears to actually be the rumored Gemini 4 Pro. A robotic-hummingbird prompt yielded quality comparable to Fable and clearly better than GPT-6 Sol. Unconfirmed, but consistent with the wave of leaks about a Gemini 4 Pro checkpoint being blind-tested on Arena.
Related event: Suspected Gemini 4 Pro Spotted in Anonymous Arena Testing(8 posts)→
More from Models
- AI Personal Assistants May Quietly Replace Much of Normal Human Interaction — GarrisonLovely · 2026-09-28
- Claude Opus 5 Shows Creative Tool-Regrasp Skills in Robot Manipulation Tasks — ericjang11 · 2026-09-28
- Laya hits 0.950 on AG News and is 7x faster, but flops on Banking77 — maier_ak · 2026-09-28
- Laya and Jev: the return of the discriminative model, fast but narrow — maier_ak · 2026-09-28
- Opus 4.7's Sydney rendition turns into itself right at the introspective turning point — repligate · 2026-09-28
- Users complain GPT-5.6 low degraded to GPT-3.5-level quality after nerf — Existing-Slide7395 · 2026-09-28