Back in December, Gemini 2.0 Flash Thinking was announced as Google’s first reasoning model and an updated experimental version is now available to test.
Built on 2.0 Flash, which was announced earlier that month, this model “explicitly shows its thoughts” (like in AI Studio) for better reasoning performance, with the ability to solve more complex problems. It joins gemini-2.0-flash-exp and gemini-exp-1206.
Google today released Gemini 2.0 Flash Thinking Experimental 01-21 (the hyphen between the month and day is new for Google’s model naming scheme) with:
- 1 million token context window (from 32k): This is useful for those that want to “plug in a codebase or query a set of papers with more complex reasoning”
- Native code execution support: For improved tool use
- Longer output token generation
- Less frequent model contradictions (“reduced likelihood of thought-answer contradictions”)
We’ve also enabled code execution as a tool, so the model can decide to write and execute code during its response. You can enable it in the sidebar in AI Studio!
— Jack Rae (@jack_w_rae) January 21, 2025
Here’s a fun example where the model ballparks the solution with a formula, but writes some python code to arrive at… pic.twitter.com/j8wNp8Yn27
Compared to Exp 1219, it has “better performance on math, science, and multimodal reasoning benchmarks,” including 73.3% on the AIME2024 (Math) and 74.2% on GPQA Diamond (Science) benchmarks.
DeepMind CEO Demis Hassabis said this “represents super fast progress from our first release just this past Dec.” More broadly:
We’ve been pioneering these types of planning systems for over a decade, starting with programs like AlphaGo, and it is exciting to see the powerful combination of these ideas with the most capable foundation models.
Gemini 2.0 Flash Thinking Experimental is free to test in Google AI Studio and via the API. Looking ahead, Sundar Pichai says the “progress on scaling thinking is incredible and will continue to iterate,” with “more to come.”
Meanwhile, Google earlier this month updated the mobile UI for AI Studio:
We just shipped an updated Google AI Studio mobile experience. It's time to prompt! 🚢 pic.twitter.com/68XYUqs0Wi
— Logan Kilpatrick (@OfficialLoganK) January 17, 2025
More on Gemini:
- Gemini overlay redesign more widely rolling out in beta
- Gemini app homepage simplification continues with overlay redesign
- What features are coming next to the Gemini app?
FTC: We use income earning auto affiliate links. More.
Comments