Google обновляет свою платформу AI с новым глубоким мышлением Gemini и революционной разработкой лекарств Translation: Google updates its AI platform with new deep thinking Gemini and revolutionary drug development

Google has upgraded its reasoning mode, Gemini 3 Deep Think. This tool is designed as a solution for complex challenges in the fields of science and engineering.

In tests, this model has outperformed OpenAI’s GPT-5.2 and Anthropic’s Claude Opus 4.6, including in the ARC-AGI-2 visual puzzles, MMMU-Pro for assessing multimodal capabilities, Elo 3455, and the “Last Exam of Humanity.”

“We enhanced Gemini 3 Deep Think in close collaboration with scientists and researchers to tackle challenging scientific problems—where tasks often lack clear boundaries or a single right solution, and where data may be incomplete,” the company stated in its blog.

Gemini 3 Deep Think exhibits remarkable results in mathematics and programming and excels in natural sciences, such as chemistry and physics. The upgraded mode addresses problems at a level comparable to that of gold medalists in international competitions.

In the CMT-Benchmark, the model scored 50.5%, confirming its deep understanding of theoretical physics.

“Beyond its advanced performance, Deep Think is focused on practical applications: it aids researchers in interpreting complex data and assists engineers in modeling physical systems through code,” Google noted.

The new Deep Think is available in the Gemini app for subscribers to Google AI Ultra and through Gemini API for select developers.

Google DeepMind has also introduced the AI agent Aletheia. This model set a new record in the IMO-ProofBench Advanced benchmark by solving 91.9% of the problems, which is considered one of the most challenging in mathematics.

This neural network is built on Gemini Deep Think and includes a verification module that detects errors in solution drafts and initiates an iterative process for refinement.

A key feature of the agent is its ability to recognize when a problem cannot be solved, significantly saving researchers’ time.

Aletheia leverages Google Search to navigate complex scientific materials, reducing the likelihood of using false references and computational errors when working with scientific content.

Among the model’s achievements:

DeepMind emphasized that Aletheia’s success confirms the relevance of scaling laws: in proof-based mathematics, quality continues to improve through the effective deployment of agents.

DeepMind’s subsidiary, Isomorphic Labs, has also launched the IsoDDE engine for drug design. In challenging tests, this new tool surpassed AlphaFold 3 in prediction accuracy by twofold.

AlphaFold 3 marked a significant breakthrough, as it could predict the three-dimensional structures of proteins and their interactions with molecules. IsoDDE, however, demonstrates a whole new level:

“IsoDDE offers a scalable foundation for AI-driven drug design, providing the prediction accuracy necessary to work with new biological systems with unprecedented reliability,” the company stated in its blog.

Recall that in July 2022, the AlphaFold algorithm predicted the structures of over 200 million proteins, accounting for nearly all known compounds discovered in plants, bacteria, and animals.