Google introduced Gemini 4 Argon, the first model in its Gemini 4 family, which appeared in limited form on September 30, 2026, and opened to everyone on October 13. Google AI Pro or Ultra subscribers can find it in the Gemini app and in Search’s AI Mode, and developers can call it through the API under the model name gemini-4-argon. Argon replaced the teased “Gemini 3.5 Pro,” which never arrived, as Google skipped a version and moved straight to the next generation. Google has not officially explained the name; the article says the calm, stable gas matches the pitch of less flash and more steady work. Google shared internal examples: Argon helped refactor around six million lines of web page code in Chrome to reduce crashes, a project that would normally take months and was done in weeks; an Argon-powered agent replaced about 32,000 lines of low-level video decoding code with safer Rust while testing its own work; one demo fed Argon more than 80,000 Mars weather logs and asked for a slide deck; and Google trained Argon to find, check, and fix software vulnerabilities, with a special version aimed at approved security defenders that watches the model’s reasoning to stop it from going off-track. Some early reviewers said its writing, knowledge, and long-text handling feel stronger than its coding and agent skills.