Gemini 4 Argon: Built to solve your hardest problems

Google has released Gemini 4 Argon, which it calls its latest frontier model, in a post dated October 13, 2026, by Koray Kavukcuoglu, SVP at Google DeepMind and the company's Chief AI Architect. The model is available globally through the Gemini API in AI Studio, in Google Antigravity for AI Ultra subscribers, and in the Gemini app and AI Mode in Search for Ultra and Pro subscribers. Introductory pricing is $2 per million input tokens and $10 per million output tokens, with a 1 million unified input-output token limit via the API.

Google says Argon ranks first on the Vals Index, which covers sectors such as finance, coding, legal and tax. Internally, the company reports agentic code submissions up more than 35% and rollbacks down nearly a third. It also says Argon helped the Chrome team refactor code from a roughly 6 million line legacy folder, and sped Gemma 4 inference by 2.85x.

Science claims include formally verified quantum arcsin circuits with spacetime costs over 100x lower than a published baseline, five novel enzyme designs confirmed active in the lab, and work with University of Maryland professor Matt Landreman on counterexamples to Grad's conjecture in fusion. Wiz used the model to find security flaws affecting government and healthcare systems.

On safety, Google reports that external bio red teaming unlocked under 1% of tested sub-tasks and that prompt injection attack success fell to 0.7% on the Gray Swan benchmark. The company is expanding its Fairwind cyber defender program and offering 10,000 scientists six months of free top-tier Antigravity access. These results are Google's own claims.