Gemini 2.5 Flash

What it is

Gemini 2.5 Flash is a model in Google's Gemini 2.5 family, offered through the Gemini API and the Gemini app. Google has said it continues to lead as the model with the best price-to-performance ratio, citing the analysis firm Artificial Analysis along with the company's own documentation for pricing ^[1]. That framing points at the kind of work it was built for: large volumes of requests where cost, latency and throughput matter, rather than the top end of the family's capability range.

How it arrived

Google released the model in preview on 17 April 2025, in the Gemini API through Google AI Studio and Vertex AI, and in a dedicated dropdown in the Gemini app ^[1]. The announcement was explicit that this was not yet a finished product: Google said the model was still in preview "before we make it generally available for full production use" ^[1].

Thinking, on or off

Google called Gemini 2.5 Flash its first fully hybrid reasoning model, meaning developers can turn thinking on or off ^[1]. The control is a thinking budget, which developers can set anywhere from 0 to 24,576 tokens; a budget of 0 gives the lowest cost and latency ^[1].

The term needs a little unpacking. In a reasoning model, "thinking" is the intermediate work the model produces for itself before it writes the answer the user sees — steps, checks, partial attempts. A thinking budget is a cap on how many tokens that hidden work may consume. Setting it high gives the model room to deliberate on hard problems; setting it to 0 removes that stage entirely, which is what makes the cheapest and fastest setting possible. In other words, the same model can be tuned across a range of cost and speed without the developer switching to a different system.

General availability and pricing

The model became generally available and stable on 17 June 2025 under the name gemini-2.5-flash, built on the 05-20 preview version ^[2][3]. Pricing changed at that point: $0.30 per 1 million input tokens, up from $0.15, and $2.50 per 1 million output tokens, down from $3.50 ^[2]. The previous price difference between thinking and non-thinking use was removed ^[2]. Google also shipped updated previews on 20 May 2025 and 25 September 2025 ^[3].

Where it sat in the lineup

For a period Gemini 2.5 Flash was the default model in the Gemini app, until Gemini 3 Flash replaced it on 17 December 2025 ^[4]. The line also produced an image variant, Gemini 2.5 Flash Image, released in August 2025 under the nickname "Nano Banana" ^[3]. Google has scheduled gemini-2.5-flash to shut down on 1 December 2026, with requests automatically routed to gemini-3.5-flash-lite ^[3]. The sequence — preview, stable release, a variant spun off for images, then replacement by a newer generation and a scheduled retirement — is a fairly complete picture of how quickly a model name moves through its working life.