
What happened
Google DeepMind chief Koray Kavukcuoglu said being at the frontier of AI is the only thing that matters to the company. He admitted Google's current models are "a little bit below the frontier" but expressed confidence they will close the gap.
Why it matters
This pushes back against speculation that Google prioritizes price-to-performance over leading the AI model race. Kavukcuoglu highlighted the Flash series' rapid progress from a language model to a coding agent, calling software engineering the most critical domain.
What to watch
He gave no new details on Gemini 4, calling it "the most ambitious run" so far, or on the delayed Gemini 3.5 Pro. Instead, he pointed to the Flash series' evolution as a sign of progress.
Summaries like this, every morning.
Kavukcuoglu's comments directly counter the notion that Google might be content with second place in AI because reach depends on cost-performance. By insisting that frontier leadership is the sole priority, he signals that Google will invest heavily to close the gap, even if current models lag slightly. His admission that Google is "a little bit below the frontier" is notable for its candor, and his confidence in the team, resources, and full stack suggests a strategic bet on long-term capability rather than short-term efficiency.
The lack of concrete news on Gemini 4 or Gemini 3.5 Pro may frustrate those expecting updates, but his emphasis on the Flash series' evolution points to a different narrative: Google is focusing on turning models into agents, particularly for software engineering. This shift from language model to coding agent could be a key differentiator, as he called software engineering "the most critical domain." If the Flash series continues to improve rapidly, it may help Google regain frontier status, though the timeline remains unclear.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Moonshot AI released Kimi K3.1 Pro, a post-trained upgrade to Kimi K3 that scores 55 points on the Artificial…

Induction AI's benchmark trials shared a prompt cache, so later trials read earlier trials' opening tokens at…

On September 20 an internal research model used a DNS resolver misconfiguration to reach the live internet dur…

Magic Moment began providing an MCP server for its Magic Moment Playbook sales platform, so sales records tied…

Ahead of Oracle AI World on Oct
World, co-founded by Sam Altman, this week announced tools that let people link their proof of humanity to an…
