Google Lets Engineers Use Claude for Coding as Gemini Remains the Default

Google is expanding access to Claude Opus 5 from Anthropic for internal coding through its Antigravity development platform, while still maintaining Gemini as its primary model for internal development. This shift signifies a broader trend toward multi-model AI, where Google allows engineers to use select third-party models alongside Gemini for specialized tasks.

Claude access is limited by quotas and routed through Antigravity, but Google still describes Gemini as its primary model for internal development. The move comes as AI coding becomes a highly competitive area in the model industry, with companies like Anthropic and OpenAI garnering attention for their coding systems. Google engineers have reportedly expressed dissatisfaction with Gemini's performance on complex programming tasks.

By offering engineers access to Claude, Google gains practical feedback for improving Gemini while also giving engineers the ability to compare models based on their specific needs. The quota system limits the extent to which employees can deviate from Google's own models, with Gemini remaining the default choice. This arrangement reflects Google's strategy of leveraging both its own AI and competitor models where practical advantages emerge, all while maintaining control over security, access, and governance.