Improved Gemini 2.5 Flash and Flash-Lite (developers.googleblog.com)

Google announced updated preview releases of Gemini 2.5 Flash and 2.5 Flash‑Lite on Google AI Studio and Vertex AI, aimed at improving quality and efficiency. The Flash‑Lite preview is available as gemini-2.5-flash-lite-preview-09-2025 and the Flash preview as gemini-2.5-flash-preview-09-2025. Early tester feedback highlights notable gains — Manus reports a ~15% improvement on long‑horizon agentic tasks and meaningful cost-efficiency, suggesting these builds are tuned for faster, more capable agent workflows and lower inference cost. Technically, these are iterative preview builds (not new stable releases) intended for testing and feedback; Google also introduced stable aliases gemini-flash-latest and gemini-flash-lite-latest that always point to the newest preview to simplify experimentation. Developers who need production stability should continue to use gemini-2.5-flash and gemini-2.5-flash-lite. Google will give two weeks’ notice before changing or deprecating versions behind the -latest aliases, and warns that rate limits, costs, and feature sets may vary between previews. Overall, this release signals incremental but practical improvements for agentic AI and cost-sensitive deployments, while letting the community influence the next stable Gemini iterations.