Google has given all of its engineers access to Anthropic's Claude Opus 5 through Antigravity, its internal development platform, per Business Insider. Until now Claude was limited to select Google DeepMind teams and high-priority projects, and engineers were otherwise expected to code with Gemini. Google said third-party models come on a quota basis for specialized uses and that "Gemini remains our primary and foundational model for internal development."
Each engineer gets a personal usage quota for Opus 5, which caps spend and keeps Gemini the primary model inside Antigravity. Coding is where the gap shows. Google's recently upgraded Flash 3.8 still trails Anthropic and OpenAI on coding, and some engineers had complained about Gemini on coding tasks, per Seoul Economic Daily. A quota is how a company admits a tool is better at one job without making it the standard. Amazon, another Anthropic investor, already lets employees use Claude and Codex.
The financial ties make this less awkward than it looks. The money circles back. Google committed up to $40 billion to Anthropic in April, $10 billion upfront and $30 billion tied to performance targets, per CNBC, and it supports a large share of Anthropic's compute expansion under that deal. Every Opus 5 session a Google engineer runs is revenue to a company Google partly owns.
Bottom Line
Google pays a rival for coding help its own model cannot yet match, and caps it per person. Enterprise teams that standardized on one vendor now have a precedent for running a second model just for code.