Google has unveiled Gemini 3, its most advanced foundation model yet, raising the bar in the rapidly accelerating AI race. The model is now available through the Gemini app and Google’s AI Search interface, with the company positioning it as its most capable large language model to date.
The release comes just seven months after Gemini 2.5 and lands days after OpenAI introduced GPT-5.1, underscoring the unprecedented pace of frontier model development. It also follows Anthropic’s Sonnet 4.5 launch two months prior, highlighting intensifying competition among the world’s leading AI labs.
A research-intensive variant, Gemini 3 Deepthink, will roll out in the coming weeks to Google AI Ultra subscribers after additional safety evaluations.
Record-breaking benchmark performance
Google says Gemini 3 delivers major improvements in reasoning.
“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 independent benchmarks seem to support this claim.
- On Humanity’s Last Exam, a test designed to evaluate broad reasoning ability, Gemini 3 scored 37.4—the highest recorded score to date, surpassing the previous record of 31.64 held by GPT-5 Pro.
- It also topped the LMArena leaderboard, a human-led evaluation measuring user satisfaction and real-world task performance.
Introducing Antigravity: Google’s new agentic coding environment
Alongside the model release, Google launched Google Antigravity, a Gemini-powered coding interface built to support agentic, multi-step development workflows.
Antigravity combines:
- a ChatGPT-style conversational prompt window,
- a command-line terminal, and
- a real-time browser pane for visualising code changes.
This setup mirrors the agentic coding features popularised by tools like Warp and Cursor 2.0, enabling developers to execute commands, inspect output, and generate or modify code in one integrated workspace.
“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 Koray Kavukcuoglu, CTO of Google DeepMind.
A new phase in the AI model race
With Gemini 3, Google is signaling its intention to stay competitive as major AI players push toward increasingly capable, reasoning-driven systems. As safety, reliability, and developer tooling become central differentiators, the release marks another step in the global race to define the next generation of AI experiences.
