News
ReleasesOVHcloud unveils Olala mini, its first in-house LLMChinaA 9B model beats a 1T model on FinQA, trained with Crusoe Managed RLReleasesThe internal metaphysics of an LLMReleasesMapping Language Model Internal Representations to Physical Laws and Scientific ReasoningResearchCan China keep rogue AI in check? Meet the start-up testing the top modelsFrontierMicrosoft aims to challenge OpenAI, Anthropic with cheaper, specialized AI modelsOpen modelsMicrosoft rolls out new first-party models and updates to agentic AI continuous improvementFrontierMicrosoft unveils AI models for coding, cybersecurity as it takes on Anthropic, OpenAIFrontierGoogle is making it easier to opt out of Gemini AI model training – here’s howFrontierOpenAI Unveils New Astra AI Model After Scrapping Prior ReleaseReleasesOVHcloud unveils Olala mini, its first in-house LLMChinaA 9B model beats a 1T model on FinQA, trained with Crusoe Managed RLReleasesThe internal metaphysics of an LLMReleasesMapping Language Model Internal Representations to Physical Laws and Scientific ReasoningResearchCan China keep rogue AI in check? Meet the start-up testing the top modelsFrontierMicrosoft aims to challenge OpenAI, Anthropic with cheaper, specialized AI modelsOpen modelsMicrosoft rolls out new first-party models and updates to agentic AI continuous improvementFrontierMicrosoft unveils AI models for coding, cybersecurity as it takes on Anthropic, OpenAIFrontierGoogle is making it easier to opt out of Gemini AI model training – here’s howFrontierOpenAI Unveils New Astra AI Model After Scrapping Prior Release
EveryLLM
Gemini 2.5 Flash

Gemini 2.5 Flash

A hybrid reasoning model from Google with adjustable thinking effort

Gemini 2.5 Flash is a Google language model that can reason through problems before answering, with a controllable thinking budget. It arrived as a preview in April 2025, built on the 2.0 Flash line, and is positioned as a cost-efficient thinking model for developers.

Share on WhatsApp

  • First Gemini model with a switch to turn reasoning on or off
  • Thinking budget adjustable from 0 to 24,576 tokens
  • Google reports strong results on LMArena Hard Prompts, behind 2.5 Pro
  • Preview released through Google AI Studio, Vertex AI and the Gemini app
  • Built on the Gemini 2.0 Flash foundation

What it is

Gemini 2.5 Flash is a language model from Google that can work through a problem in its own reasoning step before producing an answer. It is the first model in the Gemini 2.5 line to combine this thinking behaviour with a switch that lets developers turn it off. Google released an early preview through the Gemini API in Google AI Studio and Vertex AI in April 2025. The company describes it as its most cost-efficient thinking model, aimed at tasks where speed and price matter as much as answer quality.

How it came about

The model builds on Gemini 2.0 Flash, which Google describes as a popular foundation. An early version went out in preview through the Gemini API, Google AI Studio and Vertex AI, and also appeared in a dedicated dropdown in the Gemini app. Google said further improvements were planned before general availability for full production use. A later technical report on the Gemini 2.5 family was submitted in July 2025 and revised through December 2025.

Inside the model

The model can run in a thinking mode where it works through a prompt, splits complex tasks into steps and plans a response before writing anything. Developers control this with a thinking budget, a cap on how many tokens the model may spend on that internal step, ranging from 0 to 24,576 tokens. Google says the model is trained to judge how much thinking a prompt needs and will not use the full allowance if the task is simple. Setting the budget to 0 keeps the speed of 2.0 Flash while still improving performance.

What it can do

Google reports that the model performs strongly on the Hard Prompts set in LMArena, behind only Gemini 2.5 Pro. The company says it reaches metrics comparable to other leading models at a much lower cost and size, and describes it as offering the best price-to-performance ratio in its line. Reasoning quality improves as the thinking budget grows. Google gives examples of tasks needing different levels of reasoning, from translating a short phrase to solving a beam stress problem or writing a spreadsheet evaluator.

How to use it

The preview is available through the Gemini API in Google AI Studio and Vertex AI, and in the Gemini app. Developers set the thinking budget through a parameter in the API or a slider in the two cloud consoles. Google published API references, thinking guides and code examples in its Gemini Cookbook.

Why it matters

Google positions the model as part of its cost-to-quality frontier, adding a cheaper option alongside Gemini 2.5 Pro. The company says it continues to improve the model ahead of general availability.

Questions people ask about Gemini 2.5 Flash

What is Gemini 2.5 Flash?

It is a Google language model that can reason through a problem before answering. Developers can control how much thinking it does, balancing answer quality against cost and speed. Google released it in preview in April 2025.

How does the thinking budget work?

It caps the number of tokens the model may spend on its internal reasoning step, from 0 up to 24,576. Google says the model judges how much thinking a prompt needs and will not use the whole allowance for simple tasks. Setting it to 0 keeps speeds close to 2.0 Flash.

How can I use it?

Google offers the preview through the Gemini API in Google AI Studio and Vertex AI, and in a dropdown in the Gemini app. The thinking budget is set either with an API parameter or a slider in the cloud consoles.

How does it compare with other Gemini models?

Google says it reaches metrics comparable to other leading models at a fraction of the cost and size. On the LMArena Hard Prompts set, the company reports it ranks second only to Gemini 2.5 Pro.

Good reading and sources

    Where this page comes from

    There is no English Wikipedia article on Gemini 2.5 Flash yet, so the facts on this page come from the sources listed above: developers.googleblog.com/en/start-building-with-gemini-25-flash/, arxiv.org/abs/2507.06261. Some are the subject's own pages, so its claims about itself are reported as its claims. EveryLLM wrote the text in its own words, with AI assistance, and compares it with the sources to make sure no passage is copied.

    It may contain mistakes, and figures change over time. Nothing here is investment or buying advice. If something is wrong, tell us. Last refreshed 2026-11-14.

    The weekly digest

    Every Monday: the model of the day, new pages on EveryLLM, and the conferences and anniversaries of the week ahead.