Gemini App Restores 2.0 Flash Thinking (Experimental) On Android, IOS

Over the weekend, all Gemini app users on Android and iOS lost access to 2.0 Flash Thinking (experimental). This was just a bug, and the model has been restored today.

Google’s first reasoning model was released in December and updated for developers (Google AI Studio) in January before coming to the Gemini app in February as an “Experimental” offering. It was last updated in March with “better efficiency and speed,” apps access (with the secondary model removed), and document upload. Meanwhile, Gemini Advanced subscribers got to use it with a 1 million token context window.

On Saturday, all mobile users saw it disappear from the model picker, even as it remained available on gemini.google.com. As of Monday morning (PT), access is rolling back out. You might see a “New” badge attached, but there hasn’t been any model update. (In fact, it remains gemini-2.0-flash-thinking-exp-01-21 in AI Studio.)

Gemini 2.0 Flash remains the only stable model in the app, with 2.0 Flash Thinking (experimental) being the next closest one. It remains to be seen when or whether Google will make it generally available in light of Gemini 2.5 where thinking is built-in natively. Gemini 2.0 Flash Thinking (experimental) currently powers Deep Research.