#AI Studio

'The Gemini 2.5 Flash and Flash‑Lite preview updates sharpen tool use and multi‑pass reasoning while Flash‑Lite tests as the fastest proprietary model with ~50% fewer output tokens; validate on your workloads before switching to rolling aliases.'

Google has launched gemini-embedding-001, a state-of-the-art multilingual text embedding model with flexible dimensionality, leading benchmark scores, and broad integration support for AI developers.