Google spends billions of dollars every year telling the world that Gemini is the smartest AI around. So it raised eyebrows this week when a report revealed that inside Google's own offices, every engineer can now use the competition's best coding model. Google engineers officially have access to Anthropic's Claude — and the company confirmed it on the record.
01.What just changed inside Google
Business Insider reported, and a Google spokesperson confirmed, that Claude Opus 5 — one of Anthropic's strongest coding models — is now available to engineers across the company through Antigravity, Google's internal agentic development platform. That's a real policy reversal. For years, most Google employees were barred from using outside coding tools like Claude Code or OpenAI's Codex, and were expected to build with Gemini instead. Exceptions existed only for select Google DeepMind teams and a handful of high-priority projects.
And no, engineers can't just install Claude Code and go wild. The access runs strictly through Antigravity, Google's own tooling — so even when Google's engineers borrow a rival's brain, they do it inside Google's house.
02.The catch: quotas, and Gemini still comes first
Google was careful to frame the change as a supplement, not a surrender. "Gemini remains our primary and foundational model for internal development, with third-party models available on a quota to support specialized use cases," the spokesperson said. Translation: every engineer gets an individual usage allowance for Claude, and Gemini stays the default. The company hasn't disclosed how big those quotas are — which, as more than one observer has noted, makes this read more like a pressure valve than a strategic pivot.
03.Why Google blinked
According to the reporting, the change followed sustained internal frustration with Gemini's performance on complex coding tasks — even as Google pushed staff to prove they were using AI to be more productive. That's an awkward combination: mandate AI-assisted work, but ration the tools that do it best. Google's Flash-tier models have improved, but the company still trails Anthropic and OpenAI on coding benchmarks and developer sentiment.
There's a fun wrinkle here: Google is both Anthropic's rival and one of its biggest backers, with plans announced earlier this year to invest up to $40 billion in the Claude maker. Amazon, another Anthropic investor, made a similar call earlier this year and let its own employees use Claude and Codex after internal complaints. The lesson seems to be that when your engineers can't ship fast enough with your own model, ideology loses to shipping.
04.The real lesson for Indian developers and teams
Here's the thing nobody's talking about: the most valuable takeaway isn't about Google or Anthropic at all. It's that even the world's richest AI lab refuses to be dogmatic about model choice — because the people doing real work know which model actually gets the job done for a given task. If you're a developer in Bengaluru, Pune or Hyderabad, the same logic applies to you: don't marry one model. Route tasks to whichever AI does them best, keep your setup model-agnostic, and manage costs deliberately. That's exactly the discipline we teach in the cost-optimization chapter of our GPT-6 Astra — Zero to Hero Guide — when to use a heavy frontier model, when a cheap fast one wins, and how to keep your bill sane in rupees while you experiment. Anthropic's own docs and launch notes are also worth a skim — for context, we covered Claude's new Docs and Slides tools in an earlier post.
And Claude is earning that trust on real work — it now helps build the next version of itself.
Key Takeaways
- Google engineers company-wide can now use Claude Opus 5, but only inside Antigravity and under per-user quotas
- Gemini remains Google's "primary and foundational" internal model — Claude is officially a supplement
- Reported reason: engineer frustration with Gemini on complex coding, plus pressure to show AI productivity
- Google plans to invest up to $40 billion in Anthropic — rival and investor at the same time
- The takeaway for Indian devs: stay model-agnostic and route each task to the AI that does it best
Frequently Asked Questions
Why is Google letting its engineers use Claude?
Google has opened Anthropic's Claude Opus 5 to engineers across the company via its internal Antigravity platform, after engineers reportedly grew frustrated with Gemini's limits on complex coding tasks.
Is Gemini being replaced at Google?
No. There are per-user quotas, and Gemini remains the default “primary and foundational” model. Google is both Anthropic's rival and its investor, with plans to put up to $40 billion into the company.
What This Means for You
If you're a student or developer in India, don't read this as gossip — read it as permission. The instinct to defend one AI tool like a cricket team is natural, but the professionals winning right now are the ones who treat models like tools in a toolbox: Claude for gnarly refactors, GPT-6 Astra for agentic workflows that browse and operate apps, lighter models for quick everyday questions. That flexibility, plus solid prompting and cost habits, is worth more than any brand loyalty. If you want that skill set laid out end to end — prompting, agents, computer use, business ideas, cost optimization and safety — our GPT-6 Astra — Zero to Hero Guide is the practical next step.