Positioning
A preview of Google’s fast, cost-efficient Gemini 2.5 Flash-Lite model, designed for high-volume, latency-sensitive workloads. It was released on September 25, 2025. (developers.googleblog.com)
Capabilities and uses
- Accepts text, images, audio and video, and produces text. Google’s Flash-Lite model card describes a 1 million-token context window and a 64K-token output limit. (storage.googleapis.com)
- Compared with the earlier stable model, Google reports improved instruction following, more concise responses, better image understanding, more accurate audio transcription and improved translation. Suitable uses include translation, classification, document summarization and multimodal analysis. (developers.googleblog.com)
- Flash-Lite belongs to Google’s hybrid reasoning family, with thinking that can be enabled or disabled. (storage.googleapis.com)
Limitations and availability
The model can hallucinate and has limitations in causal understanding, complex logical deduction and counterfactual reasoning; adherence to thinking budgets may be inconsistent. Google lists a January 2025 knowledge cutoff for Flash-Lite. (storage.googleapis.com)
The preview was shut down on March 31, 2026. Google’s deprecation page recommends gemini-3.1-flash-lite as a replacement. At launch, Google said this preview was not intended to become a new stable release. (aistudio.google.com)
Pricing
Historical USD input and output rates for this exact preview ID could not be verified from the official pricing page as checked on November 18, 2026. The current page lists prices for the separate stable gemini-2.5-flash-lite model; those rates should not be assumed to apply to this retired preview. (aistudio.google.com)
Official sources
- Launch announcement
- Gemini 2.5 Flash-Lite model card
- Gemini API deprecations
- Gemini API pricing — checked November 18, 2026.