Google has released a new artificial intelligence model called Gemini 2.0 Flash Thinking Experimental, designed to enhance reasoning capabilities in complex problem-solving tasks. The model, available through Google’s AI Studio platform, is described by the company as being optimized for multimodal understanding, reasoning, and coding across fields including programming, math, and physics.
The new model builds upon Google’s recently announced Gemini 2.0 Flash and appears to compete directly with OpenAI’s o1 reasoning model. According to Google DeepMind’s chief scientist Jeff Dean, the system is trained to use sequential thoughts to strengthen its reasoning process. A key feature of the model is its ability to show users its step-by-step reasoning through a dropdown menu, offering transparency into how it reaches conclusions.
Initial testing reveals both capabilities and limitations of the experimental model. While it can handle complex puzzles involving both visual and textual elements, results have been inconsistent on simpler tasks. The model currently supports 32,000 tokens of input and can produce 8,000 tokens per output response. According to independent analysis from LM Arena, Gemini 2.0 Flash Thinking has achieved top performance across LLM categories, though complete details about its training process, architecture, and licensing terms have not yet been released.
Sources: TechCrunch, VentureBeat