Google makes 2.0 Flash the default model in the Gemini app

What actually changed in the app

The concrete change is a default-model swap: the Gemini app — web and mobile — now runs on Gemini 2.0 Flash. Google frames it around speed and everyday tasks, citing "fast responses" and use cases like brainstorming, learning, and writing.

This is a front-end decision as much as a model release. Users didn't opt into 2.0 Flash; it became the model behind the same chat box they already used. The interface stays constant while the engine underneath is replaced.

Alongside it, image generation in the app was upgraded to the latest version of Imagen 3, which Google says produces "richer details and textures" and follows instructions more accurately. That's a second under-the-hood substitution shipped in the same post.

The overlap window for 1.5 conversations

The detail worth noticing for anyone building on top of a chat surface is how Google handled continuity. Gemini 1.5 Flash and 1.5 Pro don't disappear the moment 2.0 Flash arrives.

Gemini 1.5 Flash and 1.5 Pro will remain available for the next few weeks for you to continue your existing conversations.Montana Labs

That phrasing is specific: the old models stay reachable to finish in-flight threads, not to start new ones. It acknowledges that a live conversation is bound to the model that produced it — switch the model mid-thread and the context and behavior shift. A brief overlap lets sessions age out gracefully instead of breaking on the day of the cutover.

Tiering built into the same interface

The announcement also draws the line between free and paid inside the app. Gemini Advanced users keep a 1M token context window supporting up to 1,500 pages of file uploads, plus priority access to Deep Research and Gems.

So 2.0 Flash is the shared baseline, while context length, upload volume, and feature access are the paid differentiators. The model itself isn't the paywall here — the surrounding capabilities are.

Enterprise accounts get 2.0 Flash "in the coming days," a staggered rollout behind the consumer web and mobile apps rather than simultaneous.

The migration is the interesting part, not the benchmark

Google mentions "stronger performance across a number of key benchmarks" without naming them, so there's little to evaluate on that front from this text alone.

The transferable lesson is operational: when a chat product replaces its default model, the hard problem is the in-flight sessions, not the new capabilities. Google's answer — keep the prior models available for existing conversations for a few weeks — is a reminder that a chat front end carries state, and that state is tied to a specific model version. Any team shipping a conversational surface has to plan the same overlap, or accept that live threads will behave differently the instant the swap lands.