Update, Oct. 13: Google has since released the model as Gemini 4 Argon. After an early unveiling that gave cyber defenders first access (our coverage), it became globally available through the Gemini API, AI Studio, Antigravity and the Gemini app for paid subscribers, at an introductory $2 per million input tokens and $10 per million output tokens, Google said.
Gemini 4 has entered early post-training, and Google hopes to release it “much earlier” than the end of the year, Google DeepMind chief Koray Kavukcuoglu said at The Information’s AI Agenda Live Summit this week. He gave no release date.
“Our intention is to, like — as soon as possible — to release an early post-training output because we see the results and we are excited,” Kavukcuoglu told The Information, adding that Google would “continue the fast-paced iterations,” The Verge reported. Post-training is the stage where a base model’s behavior is refined before release. For Gemini 4 that includes safety testing and guardrails, and Google engineers are already using the model internally to power its Antigravity coding tool, 9to5Google reported.
It was Kavukcuoglu’s first media appearance since he took over Google DeepMind in August. He had been the lab’s technology chief; Demis Hassabis stepped down as CEO and now chairs the unit and serves as Alphabet’s chief scientist, The Decoder reported.
A long gap between flagships
Gemini 4 would be Google’s first new flagship since Gemini 3 Pro in November 2025, followed by Gemini 3.1 Pro in February. Sundar Pichai announced Gemini 3.5 Pro at I/O in May for a June release, but it never shipped. Kavukcuoglu said Google “took a little bit of a step back” to focus on faster Flash models, the most recent being Gemini 3.8 Flash, and didn’t say whether 3.5 Pro has been canceled.
Google first named Gemini 4 in July, calling it the company’s “most ambitious pre-training run yet.” On Alphabet’s second-quarter earnings call, Pichai said: “We are now training Gemini 4, and we’re being very ambitious with it. I am very excited by the progress I’m seeing internally on Gemini 4, and I’m confident that people will be pleased when we are putting it outside,” 9to5Google reported. He also said Google was targeting a near-monthly cadence for Flash models.
Rivals have moved faster. Anthropic launched Claude Opus 5.5 on Sept. 22, and the same day OpenAI cut GPT-6 prices in half with its Sol and Luna models, The Next Web reported.
Asked whether Google had fallen behind, Kavukcuoglu said he had “the utmost trust in the team.” On AGI, he said, “the conversation is more about are we able to build intelligent agents that we can trust.”


WebProNews is an iEntry Publication