Google DeepMind: Gemini 4 model is in the fine-tuning phase, release planned earlier than end of year
Google DeepMind head Koray Kavukcuoglu said that the Gemini 4 model is in the fine-tuning phase and the company plans to release it much earlier than by the end of the year. He previously said that leadership in frontier AI is the company's sole priority; the Gemini 3.5 Pro model, despite being announced, was never released.
Koray Kavukcuoglu, the new head of Google DeepMind, said in his first public appearance since taking office (reported by The Information, picked up by The Verge and The Decoder) that the Gemini 4 model is now in the fine-tuning (post-training) phase, and according to him, the company plans to release it "much earlier" than by the end of the year, not at the end of the year as previously assumed. According to Kavukcuoglu, the company wants to release an "early post-training output" as soon as possible and then continue with rapid iterations. He also said the company is still working on safety measures and testing; internally, engineers are already using the model for the Antigravity coding tool.
Kavukcuoglu also reiterated earlier statements that leadership in frontier AI is the sole priority for Google DeepMind: "There's nothing else that matters to us more than being at the frontier," he said, adding that he is "100% certain" the company will manage to maintain its frontier position. He acknowledged that the company's current models are "slightly below frontier level," but said the company has the team, resources, and technological foundation to catch up.
The company has not released a new flagship model since Gemini 3 in November 2025 – in the meantime, only a Gemini 3.1 update was released in February, along with several smaller models in the Flash series. The Gemini 3.5 Pro model, announced by Google CEO Sundar Pichai in May 2026 with a planned release in June, never came out; according to the Financial Times, the project missed three deadlines. Kavukcuoglu neither officially confirmed the model's status nor denied that it had been cancelled, only saying that the company "took a step back" to focus on learning speed.
Meanwhile, according to sources, OpenAI (the GPT-6 series) and Anthropic (the Mythos/Fable series) have released models that are more powerful than the current Gemini models. Former Google DeepMind head Demis Hassabis stepped down from leadership in August 2026 and now serves as chairman of the unit and chief scientist of Alphabet; unlike him, Kavukcuoglu dismisses the topic of AGI as "not the right conversation" and focuses on developing trustworthy agents. According to the cited sources, the company is also facing departures of researchers to Anthropic and OpenAI.
Why it matters
Reports show that Google is currently losing pace in the competition with OpenAI and Anthropic – its last flagship model was released almost a year ago, and the announced Gemini 3.5 Pro failed to materialize. For companies and developers considering dependence on Gemini models, both the uncertain release schedule for Gemini 4 and the recent change in Google DeepMind leadership, along with the departure of key researchers to competitors, are relevant.
What was added since the original report
Verified updates
-
Gemini 3.5 Pro was never released, even though it was announced in May 2026; the project missed 3 deadlines; Demis Hassabis stepped down from leading Google DeepMind in August 2026; Kavukcuoglu replaced him; Kavukcuoglu explicitly rejects AGI as a priority, calling it 'not the right conversation'; Google DeepMind operates under competitive pressure as a 'Gemini factory' (a model-production facility); Gemini 4 is in the 'early phase of post-training', with release planned 'much earlier' than by the end of the year
- Gemini 3.5 Pro was never released, even though it was announced in May 2026; the project missed 3 deadlines
- Demis Hassabis stepped down from leading Google DeepMind in August 2026; Kavukcuoglu replaced him
- Kavukcuoglu explicitly rejects AGI as a priority, calling it 'not the right conversation'
- Google DeepMind operates under competitive pressure as a 'Gemini factory' (a model-production facility)
- Gemini 4 is in the 'early phase of post-training', with release planned 'much earlier' than by the end of the year
-
Gemini 4 is in the refinement phase of development; The planned release is as soon as possible, sooner within the year, not at the end of the year; Google is lagging behind OpenAI GPT-6 and Anthropic Mythos; The last flagship model, Gemini 3, was released in November 2025; Kavukcuoglu wants to process and release an early post-training output
- Gemini 4 is in the refinement phase of development
- The planned release is as soon as possible, sooner within the year, not at the end of the year
- Google is lagging behind OpenAI GPT-6 and Anthropic Mythos
- The last flagship model, Gemini 3, was released in November 2025
- Kavukcuoglu wants to process and release an early post-training output
Two audiences, two different impacts
What this means
For individuals
Developers and users relying on Gemini models should not plan around a specific release date for Gemini 4 – the company only talks about trying to release it "as soon as possible," without a binding deadline.
For a business
Companies choosing AI model vendors should bear in mind that Google has not had a confirmed flagship model since November 2025, has repeatedly pushed back deadlines (Gemini 3.5 Pro has not been released after three missed deadlines), and is experiencing departures of key people to Anthropic and OpenAI.
StrategyCheck the original
Event sources
confirmed by 2 independent sources · 2 publishers, 2 independent. We count feeds from the same owner only once.