Google Deepmind released Gemini 4 in an early post-training phase, saying it is the company's first major update since Gemini 3 late last year. It is the company's first flagship model update since Gemini 3 late last year.
Google Deepmind reported Gemini 4 is in an early phase of post-training, measured on the reliability and behavior refinement stage. That compares with Gemini 3.5 Pro, which was announced for a June release but never arrived.
Gemini 4 is built on Google's internal research and targets a wide range of applications, including coding tools like Antigravity. Availability begins with an early post-training output, initially for internal use and select developers.
"Our intention is to [...] as soon as possible [...] release an early post-training output because we see the results and we are excited," said Koray Kavukcuoglu, former CTO and now head of Google Deepmind. He hopes to release it "much earlier" than the end of the year.
The announcement follows a series of delays and missed deadlines for Gemini 3.5 Pro, which the Financial Times reported missed three deadlines. The source itself frames the significance as a shift in focus from AGI to product development.
Google Deepmind did not say whether Gemini 3.5 Pro has been officially canceled, and whether Gemini 4 can repeat the success of Gemini 3 remains an open question. The source says the model work is also supposed to help Google's chip business.
Source: thedecoder
