Overview
- Google began a global paid rollout of Gemini 4 Argon on Tuesday, making the model available through the Gemini API, the Gemini app for AI Pro and AI Ultra subscribers, Antigravity for AI Ultra users, AI Studio, and AI Mode in Search.
- Argon uses a unified 1 million token budget that counts both input and output, which can limit very long exchanges, and Google set introductory API pricing at $2 per 1M input tokens and $10 per 1M output tokens through November 30 before those rates double.
- Google cites benchmark wins and internal gains—Argon tops the Vals Index and scored 87% on ERQA while internal Antigravity use raised agentic code submissions by ~35% and cut rollbacks by nearly a third—but independent tests and some employee reports show mixed results on coding and task variability.
- Safety and specialist access are layered: Google expanded its Fairwind program to give vetted cyber defenders an unguarded Argon for defensive work, and for public users it applies monitoring of agents' reasoning, prompt‑injection defenses, and other guardrails.
- The launch follows months of delays and leadership changes at DeepMind, and it places Google back into direct competition with Anthropic, OpenAI, and others while offering 10,000 scientists temporary free Antigravity access to accelerate research use and external evaluation.