Gemini's new history-crawling model is rolling out to all with Deep Research in tow

Summary

• Gemini is expanding access to its advanced features, including Flash Thinking and Deep Research, to free users, while also offering an expanded context window for paying users.
• A new Personalization model has been launched, leveraging user search history to tailor responses, with plans to integrate more Google services like YouTube and Google Photos.
• Gemini's Deep Research is now powered by Flash Thinking, providing greater transparency with detailed reasoning, and app integrations like Calendar and Maps will soon utilize Flash Thinking for more complex multi-app actions.

Google today expanded the availability of several Gemini features that were previously locked behind an Advanced subscription, all while rolling out a new model that had previously only been spotted in passing.

The updates, including the previously-leaked Personalization model, are rolling out now. For me, they've begun appearing on the web version of Gemini. Availability on the mobile app often lags behind by a few days, so keep checking regularly.

Related

This upcoming Gemini model could tap into your search history for personalized replies

The model will aptly be named 'Personalization'

Posts

For starters, Google's Gemini 2.0 Flash Thinking Experimental model, which was first unveiled in early February, now supports file uploads. Additionally, even though the model is available to non-paying users, Google is rolling out an expanded 1 million token context window for users that pay for Gemini Advanced.

For those unaware, Flash Thinking allows users to gain a deeper understanding of Gemini's thought process and reasoning, complete with all the steps the AI model takes to respond with an answer to your query. While not beneficial for all users, Flash Thinking does offer a greater level of transparency than other models.

Elsewhere, Gemini's Deep Research model, which first started rolling out in December 2024, is now being upgraded with Gemini 2.0 Flash Thinking Experimental at its core. For reference, in its initial implementation, Deep Research helped users...well... deeply research the topic at hand. It goes through several reliable websites to formulate an answer to your query. Now, with Gemini 2.0 Flash Thinking powering the model, Deep Research does all of the same while also laying down its thought process and reasoning.

Non-paying users can use Deep Research for five detailed reports per month.

Multi-app Gemini requests are coming soon

Late last week, some users in beta started gaining access to the then-unannounced Personalization model. Google has made the new model official today, and as its name suggests, it can answer queries in a manner that is tailored to your needs. The new model does this by looking into your search history — only if you give it prior permission, though.

The model is powered by Gemini 2.0 Flash Thinking Experimental, and at first, it will only be able to access your search history. Down the line, Google aims to extend the integration to other services like YouTube and Google Photos.

"Your Search history can help Gemini Apps understand you and your interests, and give you more personalized and helpful responses," reads the model's description.

Further, Google Photos will soon be an integrated Gemini app, which means you'll be able to directly probe it about the media in your gallery. Existing app integrations like Calendar, Notes, Maps, Tasks, and more, on the other hand, will soon be powered by Gemini 2.0 Flash Thinking Experimental, which should allow users to request actions that involve multiple apps. "Look up an easy cookie recipe on YouTube, add the ingredients to my shopping list and find me grocery stores that are still open nearby," for example.