Gemini 2.5 Flash (May 20 update)
2025-05-20 · Publisher : Google · Domain : Artificial Intelligence
At Google I/O, Google releases an improved version of Gemini 2.5 Flash, with better performance on reasoning, multimodal, coding and long context, while using 20 to 30% fewer tokens.
What changes
- Improvement on benchmarks for reasoning, multimodality, coding and long context.
- Token consumption reduced by 20 to 30% in Google's evaluations.
- Available to everyone in the Gemini app; general availability announced for early June.
- New in the 2.5 family: reasoning summaries (thought summaries), extended thinking budgets for 2.5 Pro, MCP support in the API and SDK.
- Deep Think, an enhanced reasoning mode for 2.5 Pro, presented as being in testing.
### Specifications - Identifier: `gemini-2.5-flash-preview-05-20` (shut down on November 18, 2025). - Availability: Google AI Studio, Vertex AI, Gemini app. - The stable model `gemini-2.5-flash` (June 17) corresponds to this version. ### Pricing See the "Gemini 2.5 Flash (general availability)" entry: the stable version, identical to this preview, is priced at 0.30 USD input and 2.50 USD output per million tokens. ### Benchmarks No numerical score in the text of the post for Flash. For 2.5 Pro Deep Think: 84.0% on MMMU (score published by the vendor). ### Key takeaways An intermediate step toward the stable version of Flash: the model is already the one that will be deployed in production a few weeks later.
Links
Tags : LLM, Google, Gemini, Flash, Google I/O