Google launches Gemini 2.0 Flash Thinking AI model that can «reason»

Google announced its first AI model focused on «reasoning», called Gemini 2.0 Flash Thinking. It is available on the platform AI Studio.

Google noted that they saw encouraging results when they increased the computation time when displaying answers. Unfortunately, the developers did not show any performance tests of their own to confirm their claims. However, according to Chatbot Arena, Gemini-2.0-Flash-Thinking is currently ranked first in all categories.

Google has listed the following options for using the Gemini 2.0 Flash Thinking model on its developer portal:

• Reflections on the most complex problems;
• Demonstration of the model’s thinking process;
• Solving complex problems in programming, mathematics, and physics.

This new model will support contexts longer than 128,000 tokens, and has a knowledge base updated through August 2024. Developers can access this new Gemini reasoning model through the Gemini API in Google AI Studio and Vertex AI.

In September, OpenAI released its model capable of «reasoning» — o1.

Source: Techcrunch