On Tuesday, Google Gemini 3 was releasedits latest and most advanced base model, which is now readily available through the Gemini app and AI search interface.
It comes just seven months later the Gemini 2.5 releasethe new model is Google’s most capable LLM to date and a direct contender for the most capable AI tool on the market. The release also comes less than a week after OpenAI released GPT 5.1 and just two months after Anthropic released Sonnet 4.5 — a reminder of the extraordinary pace of cutting-edge model development.
A more intensive version of the model, called Gemini 3 Deepthink, will also be available to Google AI Ultra subscribers in the coming weeks once it passes further rounds of security testing.
“With Gemini 3, we’re seeing this huge leap in logic,” said Tulsee Doshi, Google’s head of product for the Gemini model. “It responds with a level of depth and nuance we’ve never seen before.”
Some of this reasoning power is already captured in independent benchmarks. With a score of 37.4, the model achieved the highest score on record the benchmark of Humanity’s Last Examintended to capture common sense and know-how. The previous high score, held by the GPT-5 Pro, was 31.64. Gemini 3 also topped the leaderboard at LMARena, a human-driven benchmark that measures user satisfaction.
According to Google, the Gemini app currently has more than 650 million monthly active users and 13 million software developers have used the model as part of their workflow.
Along with the basic model, Google also launched a Gemini coding interface called Google Antigravityallowing the coding of multi-window agents similar to agent IDEs such as Warp or Cursor 2.0. Specifically, Antigravity combines a ChatGPT-style prompt window with a command-line interface and a browser window that can show the impact of changes made by the coding agent.
Techcrunch event
San Francisco
|
13-15 October 2026
“The agent can work with your processor, in your terminal, in your browser to make sure it’s helping you build that app in the best possible way,” said DeepMind director Koray Kavukcuoglu.
