Gemini 2.5 Flash and Flash-Lite (09-2025 previews)

2025-09-25 · Publisher : Google · Domain : Artificial Intelligence

Google releases improved preview versions of Gemini 2.5 Flash and Flash-Lite, with output token consumption reduced by 24% and 50%, and introduces the "-latest" aliases.

What changes

  • Flash: better tool use and agentic gains, including SWE-bench Verified from 48.9% to 54%.
  • Flash-Lite: better instruction following, more concise responses, improved audio transcription, image understanding and translation.
  • Output tokens reduced by 24% for Flash and by 50% for Flash-Lite, and therefore costs.
  • Aliases gemini-flash-latest and gemini-flash-lite-latest that always point to the most recent version.

### Specifications - Identifiers: `gemini-2.5-flash-preview-09-2025` and `gemini-2.5-flash-lite-preview-09-2025`. - Availability: Google AI Studio and Vertex AI. - These previews are not meant to become stable versions; the stable identifiers `gemini-2.5-flash` and `gemini-2.5-flash-lite` remain recommended for production. ### Pricing Pricing not disclosed in the post (same price lists as the stable versions: Flash 0.30 / 2.50 USD, Flash-Lite 0.10 / 0.40 USD per million tokens, according to the current pricing page). ### Benchmarks | Model | Benchmark | Score | |---|---|---| | Gemini 2.5 Flash (09-2025) | SWE-bench Verified | 54% (48.9% before) | Score published by the vendor. ### Key takeaways A fast preview cycle: the gains are mainly in token efficiency and in agentic capabilities.

Links

Tags : LLM, Google, Gemini, Flash, Flash-Lite