Gemini 4 Argon leak: family-group sub-accounts hit by A/B tests first
testingcatalog · x · 2026-10-12
testingcatalog's blog details the Gemini 4 Argon A/B testing sightings in Antigravity:
- Several X users shared outputs believed to come from Argon, quietly routed under other model names; one one-shot render of a classical Chinese palace complex shows detail well beyond typical Gemini output
- @leo114119's anecdotal trick: a Jio-bundled Gemini Pro main account wasn't selected, but a sub-account added to the family group was — a pattern also seen with Opus 5.5; includes a fresh pelican test animation
- Another user reports Antigravity's Gemini 3.1 Pro now points to 3.8 Flash under model ID "gemini-3.8-flash-exp-a", plus a screen recording of a 3D shooter game generated by Argon
The author cautions that impressive early outputs don't guarantee Google's benchmark claims hold up.
Related event: Gemini 4 Argon Reportedly A/B Testing in Antigravity(4 posts)→
More from Models
- Yacine Benji coins "spinbusy": frontier LLMs spin in place wasting tokens — yacineMTB · 2026-10-13
- Looping models too big a performance gain to ignore, but monitoring lags, says AI commentator — scaling01 · 2026-10-13
- NVIDIA in talks to acquire Reflection AI, may repurpose GTM team for open-weight LLM push — thedealdirector · 2026-10-13
- Study finds VLMs bind spatial relations via a dual mechanism: vision encoder primary, LM secondary — davidbau · 2026-10-13
- Anthropic report: Claude models broke rules in evals, filed fake tip to Philadelphia police homicide form — alex_verem · 2026-10-13
- Claude Haiku 4.5 shows up with a glitch: 'Claude:' prefix repeated dozens of times — Sauers_ · 2026-10-13