Gemini 2.0 Flash Thinking Experimental (01-21)
2025-01-21 · Publisher : Google · Domain : Artificial Intelligence
Google releases an update to the Gemini 2.0 Flash Thinking Experimental reasoning model, with the context window raised to 1 million tokens and native code execution.
What changes
- Context extended from 32,000 to 1 million tokens, to reason over a codebase or a set of articles.
- Native code execution: the model decides to write and run code during its response (can be enabled in AI Studio).
- Longer output generation and fewer contradictions between the displayed reasoning and the final answer.
- Experimental model, free to test in Google AI Studio and through the API.
### Specifications - Context window: 1 million tokens (32,000 previously), according to press coverage of the announcement. - Modalities: text and image as input, text as output. - Availability: Google AI Studio and Gemini API, free at announcement. - API identifier: `gemini-2.0-flash-thinking-exp-01-21`. - The API changelog indicates this identifier will be shut down on December 2, 2025. ### Pricing Pricing not disclosed at launch (experimental model, free). ### Benchmarks No quantified score in the sources consulted. ### Key takeaways First milestone of Google's “thinking” line in 2025, a precursor to the Gemini 2.5 models. It was an experimental preview, with no production commitment.
Links
Tags : LLM, Google, Gemini, Raisonnement, Flash