Submit a comment
The AI model with built-in Google Gemini 2.5 Pro thinking capabilities is free with every account and currently has an unlimited trial period.
Just a few days after its launch, Google yesterday announced that it was "rolling out the Gemini 2.5 Pro beta to all Gemini users." This was a surprising move because when it was introduced on March 25, the service was only available to paid Gemini Advanced subscribers, with the lowest package, Google One AI Premium, costing $19.99/month in the US.
Gemini 2.5 Pro is currently available on the web and will soon be available on Android and iOS apps.
Google says it wants to “get the smartest AI models into the hands of as many people as possible, as soon as possible.” The Gemini 2.5 Pro is currently at the top of the LMArena leaderboard, and it’s also leading in both math and science benchmarks. Google says it’s working on improving the model’s coding performance as well.
Billed by Google as its “smartest AI model yet,” the Gemini 2.5 Pro integrates thinking capabilities rather than providing requests as a separate variant. It’s still in beta, but already supports a number of features like app manipulation, extensions, file uploads, and even Canvas.
Models like the Gemini 2.5 Pro can reason on their own, analyzing the logical steps in their “thinking” before giving a response. The goal is to improve performance and accuracy, especially when dealing with complex problems.
Google's new AI model also scored 18.8% on the "Humanity's Last Exam" dataset, a test designed by hundreds of experts to assess the limits of human knowledge and reasoning. This is considered a very good result compared to models that do not need to use support tools.
According to9to5google, Gemini 2.5 Pro excels at creating visually appealing web applications, agent code applications, and converting and editing source code. On the SWE-Bench Verified scale, Gemini 2.5 Pro also scored 63.8% when using a custom agent setup.
TH (according to VnExpress)