Gemini 4 Argon spotted with 900K context option as Google tests Carbon internally
heypearlai · x · 2026-10-12
Gemini 4 Argon references surfaced in Antigravity with 256K/512K/900K context options (1.3x/1.8x quota for the larger tiers), plus new low/medium/high thinking effort settings on Gemini Web. Business Insider reports Google employees are already testing further Gemini 4 checkpoints internally: Barium-B is the checkpoint for the public Argon release, while a separate Carbon checkpoint reportedly feels comparable to Claude Opus 5.5 for coding.
More from Models
- Opus 5.5 is charming and brilliant but prone to self-erasure, user observes — repligate · 2026-10-12
- Dev reports Vertex AI Gemini injecting hidden developer prompt that blocks romantic RP — zxcshiro · 2026-10-12
- Screenshot shows Gemini UI set to Pro High while the model itself says it's 3.8 Flash — AssociateFirst6296 · 2026-10-12
- Matthew Berman: Claiming a model runs on a machine is meaningless without stating tokens/s — MatthewBerman · 2026-10-12
- OpenAI's unreleased model cracked most of a famous math list in one prompt, ~3 hours of thinking each — Puzzleheaded-King584 · 2026-10-12
- GPT-6 Astra nails 72-hour agent tasks with 63% on-time rate, crushing Fable 5.1's 4% — maksym_andr · 2026-10-12