The Google Gemini logo (illustrative). Image: Google / Wikimedia Commons, Public domain, cropped
Gemini 4 is nearly done, and Google wants it out fast. Koray Kavukcuoglu, who took over Google DeepMind in August, says the model is now in post-training and should launch “much earlier” than the end of the year, 9to5Google reports. “Our intention is to roll out an early post‑training version as soon as possible, because we’ve already seen promising results and are very excited,” he said.
What he said
Kavukcuoglu was speaking at The Information’s AI Agenda Live summit this week, his first public appearance since taking the top job. Post-training is the last big stage before launch: the base model is already built, and the team is now tuning it to be reliable, useful and safe. He didn’t give a date.
A few details emerged:
- It’s already in use inside Google. According to 9to5Google, the company is testing Gemini 4 internally to power Antigravity, its AI coding tool, while safety testing and guardrails work continues.
- An early version comes first. Kavukcuoglu’s wording suggests Google may ship an early post-training release rather than wait for a finished product.
- He isn’t worried about falling behind. Asked whether Google has lost ground, he said: “I have the utmost trust in the team,” adding that “it’s a certainty that we are always gonna be at the frontier.”
Why Google is in a hurry
Its rivals have had a big September. Anthropic released Claude Fable 5.1 on September 1, and OpenAI followed with GPT-6 Astra on September 3. Google’s most recent headline model is Gemini 3.8 Flash; a planned Gemini 3.5 Pro was teased and then shelved, with Google taking “a little bit of a step back” to focus on its faster Flash models, per 9to5Google.
Investors are nervous too. Alphabet shares fell 3.8% on Wednesday, their worst day in over a month, amid worries about Meta’s Muse agent climbing the U.S. app charts, Stocktwits reports. They edged up 0.6% in overnight trading after Kavukcuoglu’s comments.
‘Not the right conversation’
The bigger shift may be in tone. DeepMind was founded to build artificial general intelligence, and Demis Hassabis, who stepped down as CEO in August and now chairs the unit as Alphabet’s chief scientist, wrote in his farewell memo: “I’ve been working towards AGI my whole life and now, like many of you, I feel it is close at hand.”
Kavukcuoglu brushed that framing aside, according to The Decoder:
The conversation of [whether] we achieved AGI or not is not the right conversation. The conversation is more about are we able to build intelligent agents that we can trust?
Koray Kavukcuoglu, head of Google DeepMind
Why it matters
Google has spent most of September watching OpenAI and Anthropic set the pace, and an early Gemini 4 is its chance to reset the race before the year ends. Kavukcuoglu’s message is also a telling shift for the lab behind AlphaGo and AlphaFold: under its new boss, DeepMind is judged less on how close it gets to AGI and more on shipping products people can trust.
Sources: 9to5Google, The Decoder, Stocktwits via Yahoo Finance, Benzinga
Related
