At Google I/O 2025, the company unveiled an upgraded reasoning mode for its Gemini AI, named Deep Think, aimed at enhancing performance. This new model can consider multiple answers before responding, leading to better benchmark scores. Google has also improved its budget Gemini 2.5 Flash model for tasks like coding and reasoning, set to be generally available in June. Additionally, the new Gemini Diffusion model promises faster output, competing against larger models. Both advancements emphasize Google's commitment to developing state-of-the-art AI technologies while ensuring safety evaluations prior to wider rollout.
[Deep Think] pushes model performance to its limits," said Demis Hassabis, head of Google DeepMind... It uses our latest cutting-edge research in thinking and reasoning, including parallel techniques.
Google says that Deep Think enabled Gemini 2.5 Pro to top LiveCodeBench, a challenging coding evaluation.
Google has introduced an update to its budget-oriented Gemini 2.5 Flash model that allows the model to perform better on tasks involving coding, multimodality, reasoning, and long context.
Gemini Diffusion, which the company claims is "very fast" - delivering output 4-5 times quicker than comparable models and rivaling the performance of models twice its size.
Collection
[
|
...
]