Sundar Pichai unveils Gemini 4 Argon, Gemini adds Claude

Gemini at Work 2026 expanded beyond a model refresh into how Google wants agents used at work. The keynote, held 2:30–3:30 a.m. Japan time on October 9, introduced Gemini 4 Argon as the new flagship and a rebuilt Gemini positioned as a "universal agent" that autonomously carries out tasks. Google's evidence for the shift was internal: agents submitted 35% more code and cut rollbacks by roughly one-third, and the Chrome team compressed refactoring periods from months to weeks.

Among the new features is a model picker in Gemini's own interface that includes Claude models (Opus) from Anthropic, alongside Auto mode and Gemini's Argon and Flash models. A demo by Adam Fischer showed a UI reminiscent of Antigravity 2.0, plus a Tasks inbox for tracking background agent progress and its "thinking process," a Skills feature for encoding company rules in markdown or prompts, connectors to existing SaaS tools, and scheduled recurring tasks such as drafting replies to unread email each morning.

Google has not disclosed a general availability or public preview date for any of this, describing the features as "coming soon" or as early partner previews, so timing and Japan-region rollout order remain open.