Google rations Opus 5 to its own engineers inside Antigravity

"Gemini remains our primary and foundational model for internal development, with third-party models available on a quota to support specialized use cases," the Google spokesperson quoted by Business Insider said ^[3]. The default stays the house model, and the exception is metered per head ^[2]. Business Insider did not report the quota size or the commercial terms ^[16].

That gap is where the competitive question lives, because the same spokesperson framed Claude as a supplement and stopped short of calling it an admission that Anthropic's models performed better ^[5]. Utilisation separates those two readings. If engineers exhaust their allowance and Google widens it, the supplement framing thins out; if usage sits well under the cap while Gemini keeps the volume, this is housekeeping in a developer platform and tells you little about which lab leads on coding. I lean to the first, mostly because the sourcing here is two employees describing frustration ^[1]. I would give the view up if Google published utilisation putting Opus 5 at a marginal share of internal coding calls.

Google has said it plans to invest up to $40 billion in Anthropic ^[8]. Whatever its engineers consume inside Antigravity, the spend moves from a backer to the lab whose coding model its own model is being measured against.

The reported reason for lifting the restriction was that the pace of engineering work mattered more than the rivalry ^[6]. The pacing argument circulating in the same week runs the other way. Bilal Chughtai, who left an AGI safety and alignment research role at Google DeepMind in July, wrote: "I earnestly believe that AI has the potential to kill us all, and that we might be running out of time to avoid this outcome" ^[9][10]. He called for coordination to avoid "this manic race between AI companies" and for development at "a speed that society can handle" ^[11]. Dario Amodei's essay of Saturday, September 12, "We Must Pace the Frontier", drew public support from Sam Altman and Elon Musk ^[12]. Anthropic's Evan Hubinger has put the odds of AI killing everyone above 10% this decade, a figure Geoffrey Hinton called "not unreasonable" ^[14]. President Donald Trump has waved the warnings away, telling reporters his worry is beating China ^[15].