Gemini 3.1 Flash-Lite

What it is

Gemini 3.1 Flash-Lite is the cheapest and fastest member of Google's Gemini 3 family of language models. Google's model page describes it as a low-latency, cost-effective multimodal model optimized for high-frequency, lightweight tasks^[5]. In practice that means a model meant to be called often and cheaply: the kind of workload where a system makes many short requests rather than a few long, difficult ones, and where price per call and response speed matter more than headroom on hard problems.

It is multimodal on the input side. It accepts text, image, video, audio and PDF files and produces text output^[5]. Its input context window is 1,048,576 tokens, and it can generate up to 65,536 tokens in a single response^[5]. The model page lists function calling, caching and the Batch API as supported features^[5]. Function calling lets a program hand the model a set of tools it may invoke; caching stores repeated prompt prefixes so they are not reprocessed and re-billed at full price on every call; the Batch API accepts large jobs that are processed together at a discount.

Release history

The model first appeared as a preview, gemini-3.1-flash-lite-preview, on 3 March 2026^[1][4]. The generally available version, gemini-3.1-flash-lite, was released on 7 May 2026, and the preview was shut down on 25 May 2026^[1]. Google recommended it as the replacement for gemini-2.0-flash-lite and for the 2.5 Flash-Lite preview, the earlier cheap tier it was meant to stand in for^[1]. Releasing under a preview name first and then switching to a stable identifier is the usual pattern for Gemini models: developers can try the new version early, and the preview endpoint retires once the stable one exists.

Pricing

Price is the main point of a Flash-Lite model, and the Gemini Developer API pricing lists $0.25 per 1 million input tokens for text, image and video, $0.50 per 1 million input tokens for audio, and $1.50 per 1 million output tokens^[2]. Two discounts reduce the input side further. Requests sent through the Batch API or the Flex tier cost $0.125 per 1 million input tokens for text, image and video and $0.75 per 1 million output tokens^[2]. Context caching is priced at $0.025 per 1 million input tokens for text, image and video^[2] — a tenth of the ordinary input price, which follows from the work on a cached prefix being done only once.

Where it sits in the line

The model spent only a short time at the front of Google's lineup. It was succeeded by Gemini 3.5 Flash-Lite in July 2026, a model Google said significantly outperformed it^[3], and its shutdown is scheduled for 7 May 2027^[1]. Google's deprecations page now recommends gemini-3.5-flash-lite as the replacement for gemini-3.1-flash-lite^[1].

The sequence shows how quickly the lightweight tier of the Gemini line turned over in this period: a preview in March 2026, general availability in May, a successor in July, and a retirement date set less than a year after launch. For developers the practical consequence is that a "lite" model is a moving target — code that names a specific version has to be updated on a schedule, while the role it fills, cheap high-volume inference, stays the same. Gemini 3.1 Flash-Lite is best understood as one turn of that cycle: a low-cost multimodal endpoint with a large context window, built for many small requests, and replaced within months by a faster and more capable model of the same kind.