New DeepMind Chief: Google Will Not Give Up Frontier AI

Google DeepMind's new chief AI architect Koray Kavukcuoglu, in interviews with Logan Kilpatrick and Saboo Shubham, directly addressed rumors that Google is abandoning frontier AI, affirming the company's strategy remains centered on frontier models.

Confirmed

• Kavukcuoglu said that apart from being at the frontier, nothing else matters, and he is 100% certain Google will be at the frontier (as relayed by mattslotnick, firstadopter, davidblack).
• He admitted Google's current models are slightly below the frontier, directly refuting the popular narrative that Google has given up on frontier models and settled for "good enough" ones (mattslotnick).
• The interview confirmed Gemini 3.5 Pro is still in development (firstadopter).
• The conversations also covered the path to AGI, progress on the 3.7 Flash model, and why the team focuses on frontier model development (Logan Kilpatrick, Saboo Shubham).

Why it matters

• This is a significant public statement by DeepMind's new leader, countering recent speculation about a Google strategy shift and signaling continued investment in frontier models amid competition with OpenAI and Anthropic.
• Mentions of 3.7 Flash and the still-in-development 3.5 Pro give readers an official view of the Gemini roadmap.

2026-09-01 ~ 2026-09-02 · 6 related posts

• [source] Google DeepMind lead on AGI path, 3.7 Flash progress — OfficialLoganK · 2026-09-01
• Google DeepMind reiterates commitment to SOTA: 'Nothing else is important' — davidblack · 2026-09-02
• [source] DeepMind SVP: Google won't give up on frontier AI, Gemini 3.5 Pro still in works — firstadopter · 2026-09-02
• [source] New DeepMind Chief: Google Must Stay at the Frontier, Rejects 'Good Enough' Strategy — matt_slotnick · 2026-09-02

2 near-duplicate retellings: OfficialLoganK · Saboo_Shubham_