Google DeepMind’s new chief, Koray Kavukcuoglu, is pushing back on the idea that Google is comfortable competing mainly on distribution or price-to-performance rather than frontier model leadership. According to The Decoder, Kavukcuoglu said that, for Google DeepMind, being at the frontier is the priority and that he is “100% certain” the company will get there. The notable part is the admission embedded in the confidence. The Decoder reports that Kavukcuoglu said Google’s current models are “a little bit below the frontier,” while arguing that the organization has the team, resources, and full-stack capability to close the gap. There was no matching product reveal in the item. The Decoder says Kavukcuoglu did not share concrete frontier news to support the claim. On Gemini 4, he repeated that it is Google DeepMind’s “most ambitious run” so far and said it is going well, but the outlet characterizes that as already known rather than a new update. The same applies to Gemini 3.5 Pro. The Decoder reports there was no update on that model, which it says is apparently still in development and months late. From the information provided, the story is therefore a strategy signal, not a launch. Kavukcuoglu instead pointed to the Flash series as evidence of progress. The Decoder reports that he described the jump from 3.5 through 3.6 to 3.7 as a move from a language model toward a coding agent, and said software engineering is the most critical domain where these systems need to succeed. For readers, the clean read is that Google DeepMind is publicly rejecting a “good enough plus distribution” posture. But the available evidence does not yet show whether that position has translated into a new frontier release, a benchmark lead, or a dated roadmap for Gemini 4 or Gemini 3.5 Pro. Who benefits: Too early to tell from the provided reporting. Kavukcuoglu pointed to DeepMind’s team, resources, and full-stack capability, but The Decoder did not report a concrete frontier release. Who's exposed: Google is exposed to scrutiny if its public confidence is not matched by releases or benchmarks. Teams waiting on Gemini 3.5 Pro or Gemini 4 have little new timing detail from this report.