Google has introduced 'Deep Think,' an enhanced reasoning mode for its Gemini 2.5 Pro model, at the annual I/O developer conference on May 20-21, 2025. This new feature equips the AI with advanced research techniques, enabling it to consider multiple hypotheses before formulating responses, designed for complex math and coding tasks.
Gemini 2.5 Pro with Deep Think achieved an impressive score on the 2025 USAMO and leads on LiveCodeBench. It also scored 84.0% on MMMU, testing multimodal reasoning. Google is conducting safety evaluations and gathering expert input before wider release via the Gemini API.
The company also announced improvements to Gemini 2.5 Flash, available in Google AI Studio for developers and in Vertex AI for enterprises in early June. These advancements underscore Google's focus on integrating AI across its platforms.