December 6th news, yesterdayGoogle Official launch of the new generationinference model「Gemini 3 Deep Think, a version known as the "IMO Gold Level" model, has achieved a significant leap forward from the previous generation Gemini 2.5 Deep Think, leading in many international benchmarking tests。
Gemini 3 Deep Think, using a "paradex" structure, is able to explore multiple assumptions at the same time, thus demonstrating a breakthrough in complex mathematical, scientific and logical dilemmas。
A high score of 41% was obtained in the Humanity's Last Exam (no tool-assisted) test, reaching 451% in the ARC-AGI-2 (coding) test, which is the global lead。
Meanwhile, Google announced the formation of a new elite team in Singapore, led by Chinese scientist Yi Tay, focusing on advanced reasoning, LLM/RL and Gemini。
Currently, the Gemini 3 Deep Think model is open to Ultra users in Gemini App。