Google releases Gemini 4 Argon model globally via API, Antigravity, Gemini app and AI Mode in Search at $2/$10 per million tokens

On 13 October 2026 Google made Gemini 4 Argon generally available. Developers and enterprises get it through the API in AI Studio and the Gemini platform. Google AI Ultra subscribers get it in Antigravity, and Ultra and Pro subscribers get it in the Gemini app and AI Mode in Search. Introductory pricing is $2 per million input tokens and $10 per million output tokens, with cached input 95% cheaper. Google says it ranks first on the Vals Index and cites internal results, such as 35% more agentic code submissions in Antigravity and a 2.85x Gemma 4 E2B speedup via vLLM.

This is a real, usable release with published prices, so developers and enterprises can test it against their current models straight away. The cached-input discount matters most for agent and long-context workloads that resend large prompts. Putting it in AI Mode in Search gives it consumer reach that rivals cannot easily match. Whether it is actually better or cheaper in practice for your workload is unproven. The word "introductory" also means the price may rise.

"Months of engineering work in weeks" and "outperforms other frontier models" are Google's own claims, and no independent party has verified them. The internal figures (35% more submissions, fewer rollbacks, a Chrome refactor) have no disclosed method or baseline. The linked ADK repo and xTB speedups are narrow tooling gains, not evidence of overall model quality. A first place on a single index says little about your tasks.

1. 01Independent evaluations over the next 2 to 4 weeks (the Vals Index itself, plus third-party coding and agent benchmarks) to see whether the lead holds outside Google's own reporting.
2. 02Whether the introductory $2/$10 pricing is kept, raised or given an end date, and real per-task costs once reasoning tokens and rate limits are counted.
3. 03Competitor responses (price cuts or new model releases) and developer adoption signals, such as API migration reports and usage rankings on routers like OpenRouter.