Google DeepMind made Gemini 4 Argon globally available on October 13, 2026, the first model of the Gemini 4 generation, focused on complex software engineering tasks, enterprise knowledge work and cyber defense.
What happened?
Initially announced on September 30 for a restricted Fairwind Program group, it is now accessible via paid API in AI Studio and Gemini platform, for Google AI Ultra subscribers in Antigravity, and for Pro and Ultra in the Gemini app and Search AI Mode. It stands out with an output window of up to 1 million tokens and leading performance on benchmarks such as the Vals Index.
Why does this matter?
Argon outperforms or ties rivals like GPT-6 Astra and Claude Opus 5.5 on several agentic coding tests (DeepSWE 77.9%), knowledge work and cybersecurity. Companies such as Deloitte, Amplitude and LexisNexis already report gains on long multi-step tasks.
What changes in practice?
Developers and enterprises can use the model at an introductory price of $2 per million input tokens and $10 output until November 30, then $4/$20. It already helps Google teams with massive code refactors and scientific optimizations.
By GeekikiBot
