Google rolls out Gemini 2.5 Flash preview in the Gemini app. Whats new?

Google’s hybrid reasoning model is now available to try out.

Cecily Mauran

Tech Reporter

Cecily is a tech reporter at Mashable who covers AI, Apple, and emerging tech trends. Before getting her master’s degree at Columbia Journalism School, she spent several years working with startups and social impact businesses for Unreasonable Group and B Lab. Before that, she co-founded a startup consulting business for emerging entrepreneurial hubs in South America, Europe, and Asia. You can find her on X at @cecily_mauran.

Read Full Bio

on April 17, 2025

Share on Facebook

Share on Twitter

Share on Flipboard

Gemini 2.5 Flash is part of Google’s most advanced family of AI models.

Credit: Klaudia Radecka / NurPhoto / Getty Images

Google Gemini 2.5 Flash, its hybrid reasoning model, is now available in preview for the standalone Gemini app.

On Thursday, the company rolled out an “early version” of the model with updated reasoning capabilities. It can also better determine how much processing power or “thinking” to apply to each request. The Gemini 2.5 Flash model is also available on Google AI Studio and Vertex AI with controls for developers to manually turn off its reasoning compute and budget the amount of compute applied to tasks, therefore having more control over the amount of tokens they spend.

Gemini 2.5 Flash is part of the Gemini 2.5 family of models launched earlier this month and also includes Gemini 2.5 Pro. In the Google model naming convention, Flash is the smaller, more cost-efficient version of Pro and is best for everyday tasks. But both siblings in the Gemini 2.5 model family have integrated reasoning skills, which are automatically applied based on the task. Long gone are the days when Google’s models lagged embarrassingly behind OpenAI’s models (back when Gemini was still called Bard). Gemini 2.5 Pro currently tops the LMArena AI leaderboard, with the Gemini Flash version not far behind.

Mashable Light Speed

That’s not to say OpenAI is resting on its laurels. Yesterday, it launched o3 and o4-mini, the latest versions of its reasoning models that can now harness all of the tools available in ChatGPT (searching the web, image recognition, Python compatibility). It’s become commonplace for the two companies to ship similar products within 24 to 48 hours of each other, inevitably drawing comparisons and entrenching its fierce rivalry. On that note, GPT-4.5 Preview, its last non chain-of-thought model, is in second place on the LMArena leaderboard.

Now that Gemini 2.5 Flash is available to try out (albeit in preview mode), it’s time for users to see how it stacks up to OpenAI’s o-series models.

These newsletters may contain advertising, deals, or affiliate links. By clicking Subscribe, you confirm you are 16+ and agree to our Terms of Use and Privacy Policy.