Google’s Gemini 3 AI Model Release: Key Details and Expectations

Google CEO Sundar Pichai Confirms Gemini 3 AI Model Release Later This Year

Google CEO Sundar Pichai confirmed the release of the Gemini 3 artificial intelligence model later this year. The announcement was made during the Q3 2025 earnings call.

The company reported $102.3 billion in revenue for the quarter. This milestone comes seven months after the release of the Gemini 2.5 Pro model. The Gemini 2.5 family of AI models was made available to all users globally in July.

Rumors about the Gemini 3 model have been circulating on social media. One rumor claimed the model would be released on October 22, which has been proven false. Some users speculate that the model could be tested using the names “lithiumflow” and “orionmist.” They suggest testing on the AI model ranking platform LMArena.

The rumored capabilities of Gemini 3 include improved performance in reasoning and front-end coding. Some users claim it will feature a trillion-parameter mixture of experts architecture with 15-20B tokens active per query. Other rumored features include real-time 60fps video processing and a response time of less than one second.

The exact release date of the Gemini 3 model is still unknown. Some speculate it could be released between November 15 and December 5. The model will likely be rolled out to a small user group in the US first. After that, it will expand globally.

The Gemini 2.5 models unified general-purpose and reasoning versions, adding native reasoning. Gemini 2.5 Pro outperformed leading models in several benchmarks.