Google just quietly acknowledged that Gemini is not the right tool for every coding job. The company has opened access to Anthropic’s Claude Opus 5 for all of its engineers through an internal platform called Antigravity.
What Changed
For years, Google’s internal tooling favored Gemini. That policy has now shifted. Engineers can tap Claude Opus 5 directly inside Antigravity, though Google was careful to frame the move in a way that keeps Gemini at the top of the hierarchy.
“Gemini remains our primary and foundational model for internal development, with third-party models available on a quota to support specialized use cases.”
The quota framing matters. Claude access is available, but it is not unlimited.
Why This Happened
The policy shift ends years of restrictions that had favored Gemini for internal development and reportedly eased internal tensions at the company. It reflects a pragmatic recognition that no single model dominates every coding task, according to Business Insider reporting on the change.
The move also sits inside a larger financial picture: Google has invested billions in Anthropic while simultaneously racing to match rivals, including Microsoft and OpenAI, on AI development speed. Claude Opus 5 is accessible to Google engineers via Vertex AI and Google Cloud, which are the existing infrastructure layers Anthropic’s models run on within Google’s stack.
The Operator Takeaway
For indie hackers and solopreneurs watching the model wars, this is a useful signal. Even Google DeepMind’s own engineers are now working with multiple models rather than one. The real-world verdict on which model wins specific tasks is still being written, and the answer is increasingly “it depends on the task.”
