Google Unveils Next-Gen Gemini Models: Pro and Flash Thinking Experimental

Gemini App Adds 2.0 Pro and 2.0 Flash Thinking Experimental

• Gemini 2.0 Pro Experimental: Google has introduced a new experimental model, 2.0 Pro Experimental, which is the company's best model yet for coding performance and complex prompts. It offers improved understanding and reasoning of world knowledge compared to previous models. For the Gemini API, it features a 2 million token context window, enabling comprehensive analysis of vast amounts of information. The model is available to Gemini Advanced subscribers ($19.99 per month) on both web and mobile apps.

• Gemini 2.0 Flash Thinking Experimental: This new model combines the speed and performance of 2.0 Flash with enhanced reasoning capabilities. It breaks down prompts into a series of steps, providing real-time visibility into its thought process and assumptions. A version that can interact with apps like YouTube, Search, and Google Maps is also available. Both versions are free to test in the Gemini app for all users.

• Other Announcements:

  • Gemini 2.0 Flash: Now generally available (GA) for developers using Google’s API, with pricing details provided. Image and audio output capabilities and the Multimodal Live API will follow soon.
  • Reinforcement Learning Techniques: The 2.0 family was built using new reinforcement learning techniques, leading to more accurate feedback and improved handling of sensitive prompts.
  • Gemini 2.0 Flash-Lite: A cost-efficient model optimized for large-scale text output use cases. It performs better than 1.5 Flash across most benchmarks while maintaining speed and cost efficiency.