What Google has teased about Gemini 4 - 9to5Google

What the report says

9to5Google reports that Google has begun signaling work on Gemini 4, its next major AI model family, even as Gemini 3.5 Pro has not yet launched after an expected June timing slipped. The article says the first public reference to Gemini 4 came alongside announcements for Gemini 3.6 Flash, 3.5 Flash-Lite and 3.5 Flash Cyber, with Google describing the upcoming model work as its most ambitious pre-training effort so far.

The report draws heavily on remarks by Alphabet and Google CEO Sundar Pichai during Alphabet’s Q2 2026 earnings call. Asked whether Gemini can stay competitive at the AI “frontier,” Pichai said the field is changing quickly and acknowledged areas where Google wants to improve, including coding and agentic coding. He indicated that Google is directing substantial compute and effort toward making Gemini 4 competitive with the leading models at the time it is released.

Pichai also discussed the role of Google’s Tensor Processing Units, saying the company’s top allocation priority is supporting frontier AI and AGI development. According to 9to5Google, the clearest technical hint was that Gemini 4 will require a larger base model to compete at that level, and Pichai said the model is already in training.

9to5Google adds that, based on Google’s prior model release patterns, Gemini 4 could arrive in November or December, though the article presents that as an expectation rather than a confirmed date. In the meantime, Pichai said Google plans continued updates to Gemini 3.x Flash models, with a goal of releasing new models at roughly a monthly pace.