The Gemini 3 AI model will be made available later this year, according to Google CEO Sundar Pichai
| .png) | 
| During the Q3 2025 earnings call, the Google CEO confirmed the release of Gemini 3 | 
Google CEO Sundar Pichai said on Wednesday that the Gemini 3 artificial intelligence (AI) model will be introduced later this year. The confirmation was made at his Q3 2025 earnings call, when the company's achievement of reaching the $102.3 billion (about Rs. 9.06 lakh crore) sales milestone for the quarter was the primary event. It has now been revealed that the rumored future Gemini model would launch seven months after the Gemini 2.5 Pro model. Notably, in July, the Gemini 2.5 family of AI models was made available to all users worldwide.
Google CEO Confirms Gemini 3 AI Model
There have been a lot of rumors about Gemini 3 on social media during the past few months. One such rumor, which has since been disproven, stated earlier this month that the alleged model may be introduced on October 22. Others have asserted that the corporation may be attempting covert model testing using two enigmatic models, orionmist and lithiumflow, that surfaced on the crowdsourced AI model ranking platform LMArena.Numerous users of X (previously Twitter) have revealed the potential of these rumored models, citing better front-end coding and reasoning performance. With capabilities like real-time 60fps video processing, built-in Deep Think reasoning, and a response time of less than one second (with the Flash form), some have said that it will be a trillion-parameter mixture of experts (MoE) architecture with 15-20 billion tokens active every query.
whaaat lithiumflow (rumored Gemini 3) is good (vs Grok 4)
— ❝Rijn Hartman❞ (@RijnHartman) October 21, 2025
(video comparison, who do you think did better? first dashboard is grok) pic.twitter.com/bmQIkYQDdz
Gemini 3.0 is coming today? Here’s what we know about Gemini 3.0:
— Hasnain (@aixradar) October 27, 2025
→ Trillion-parameter MoE architecture with only 15-20B active per query
→ Multi-million token context window (largest in production AI)
→ Built-in Deep Think reasoning as default mode
→ Real-time 60 FPS… pic.twitter.com/6xFIRQ35DA
Google has not disclosed the precise date of this AI model's potential public release, despite the confirmation. It may be launched between November 15 and December 5, according to several rumors, although these are only conjectures. Furthermore, these models will probably be introduced to a small user base in the United States before being made available to business customers, people worldwide, and through application programming interfaces (APIs).
Notably, by including native reasoning into the Gemini 2.5 models, the tech giant brought the general-purpose and reasoning versions of the models together. In a number of benchmarks, including GPQA Diamond, AIME 2024 and 2025, Aider Polyglot, and MMMU, Gemini 2.5 Pro was also said to perform better than top models at the time, including OpenAI's o3-mini, Grok 3 Beta, Claude 3.7 Sonnet, and DeepSeek R1.
#google #gemini3 #aimodel #newstainmentora
