Google unveils Gemini 3, its most powerful AI model yet, with major gains in reasoning and coding capabilities
Google has pushed its AI ambitions forward once again with the launch of its most advanced model to date.
- Nov 19, 2025,
- Updated Nov 19, 2025 9:02 AM IST
Google has rolled out Gemini 3, its newest foundation model, making it immediately accessible through the Gemini app and AI search. Arriving only seven months after Gemini 2.5, the update positions Google at the front of the race for the most capable AI system available today. The timing is notable, landing just days after OpenAI released GPT 5.1 and two months after Anthropic introduced Sonnet 4.5, underscoring the rapid escalation of frontier model development.
A more advanced, research-focused version known as Gemini 3 Deepthink is also on the way. It will be offered to Google AI Ultra subscribers in the coming weeks once it clears additional safety evaluations.
“With Gemini 3, we’re seeing this massive jump in reasoning,” said Tulsee Doshi, Google’s head of product for the Gemini model. “It’s responding with a level of depth and nuance that we haven’t seen before.”
Early benchmark scores support those claims. With a score of 37.4, Gemini 3 now holds the top position on the Humanity’s Last Exam benchmark, designed to measure high-level reasoning and general expertise. The previous record belonged to GPT 5 Pro at 31.64. The model has also reached the number one spot on LMArena, a user satisfaction driven leaderboard.
Google says the Gemini app currently serves more than 650 million monthly active users, and 13 million developers integrate the model into their workflow. These figures highlight how central Gemini has already become to Google’s broader AI ecosystem.
Alongside the main model release, Google introduced Antigravity, a new Gemini-powered coding interface. It brings multi-pane agentic coding to developers by merging a chat-style prompt window with a terminal and a browser view. This setup allows the coding agent to write, test and refine applications in real time.
“The agent can work with your editor, across your terminal, across your browser to make sure that it helps you build that application in the best way possible,” said DeepMind CTO Koray Kavukcuoglu.
Google’s latest launch signals an intensifying contest between AI leaders, each pushing the boundaries of reasoning, coding and agentic capability. With Gemini 3 leading new benchmark charts and Deepthink on the horizon, the company is making clear that it intends to stay at the top of this race.
For Unparalleled coverage of India's Businesses and Economy – Subscribe to Business Today Magazine
Google has rolled out Gemini 3, its newest foundation model, making it immediately accessible through the Gemini app and AI search. Arriving only seven months after Gemini 2.5, the update positions Google at the front of the race for the most capable AI system available today. The timing is notable, landing just days after OpenAI released GPT 5.1 and two months after Anthropic introduced Sonnet 4.5, underscoring the rapid escalation of frontier model development.
A more advanced, research-focused version known as Gemini 3 Deepthink is also on the way. It will be offered to Google AI Ultra subscribers in the coming weeks once it clears additional safety evaluations.
“With Gemini 3, we’re seeing this massive jump in reasoning,” said Tulsee Doshi, Google’s head of product for the Gemini model. “It’s responding with a level of depth and nuance that we haven’t seen before.”
Early benchmark scores support those claims. With a score of 37.4, Gemini 3 now holds the top position on the Humanity’s Last Exam benchmark, designed to measure high-level reasoning and general expertise. The previous record belonged to GPT 5 Pro at 31.64. The model has also reached the number one spot on LMArena, a user satisfaction driven leaderboard.
Google says the Gemini app currently serves more than 650 million monthly active users, and 13 million developers integrate the model into their workflow. These figures highlight how central Gemini has already become to Google’s broader AI ecosystem.
Alongside the main model release, Google introduced Antigravity, a new Gemini-powered coding interface. It brings multi-pane agentic coding to developers by merging a chat-style prompt window with a terminal and a browser view. This setup allows the coding agent to write, test and refine applications in real time.
“The agent can work with your editor, across your terminal, across your browser to make sure that it helps you build that application in the best way possible,” said DeepMind CTO Koray Kavukcuoglu.
Google’s latest launch signals an intensifying contest between AI leaders, each pushing the boundaries of reasoning, coding and agentic capability. With Gemini 3 leading new benchmark charts and Deepthink on the horizon, the company is making clear that it intends to stay at the top of this race.
For Unparalleled coverage of India's Businesses and Economy – Subscribe to Business Today Magazine