Google is testing a new Gemini 4 variant called Carbon that reportedly matches Anthropic's Opus 5.5 coding performance. According to internal documents and chats reviewed by Business Insider, Google deployed Carbon on its internal coding platform Jetski, where it demonstrated superior programming capabilities compared to earlier variants.
The tests reveal that Google is organizing its upcoming Gemini 4 lineup into distinct roles, positioning Argon as its frontier reasoning model alongside variants like Barium and Carbon. While early Argon versions were compared to Anthropic's Opus 5, the newer Carbon variant has drawn internal comparisons to the stronger Opus 5.5 model. This rapid iteration pace highlights how major AI labs are accelerating development cycles.
Google has not yet announced an official launch date for the Gemini 4 family, but the company is already preparing its ecosystem by updating the Gemini app and AI Studio with advanced reasoning controls. This development matters because it demonstrates the intensifying competition among leading AI labs to deliver superior coding and reasoning performance in next-generation frontier models.