Last week, Google released its highly anticipated Gemini AI model, but initially, it was only accessible to paying subscribers. In a surprising move, Google has swiftly launched the Gemini 2.5 Pro (Experimental) version, now available for free users as well. This means that during your next visit to Gemini, you can explore most of the new features without needing a Gemini Advanced subscription.
The Gemini 2.5 branch is set to replace the previous version, Gemini 2.0, which was only introduced in late 2024. This new model supports simulated reasoning, a feature that all of Google’s AI models will incorporate in the future. This innovative approach helps to minimize common mistakes that AI models have made historically, enhancing overall performance.
Users have expressed enthusiasm for Gemini 2.5's capabilities, as it has quickly ascended to the top of the LMSYS Chatbot arena leaderboard. Google states that the Gemini 2.5 Pro (Experimental) version is readily available for free users to test on the web. To use it, simply select the model from the drop-down menu, enter your prompt, and observe the model's reasoning in action.
While the free tier allows access to this new model, it’s important to note that it does not include all advanced features. For instance, users cannot upload files to Gemini without a paid account, which can hinder the ability to utilize the model's extensive context window. Although the free version won’t offer the full 1 million-token window, Google has indicated that it will have a lower limit, specifics of which have yet to be disclosed.
Google has established a framework for usage of the new model to encourage subscriptions. As a result, users will experience rate limits when using the Gemini 2.5 Pro (Experimental). Although Google has not provided clear details about these limits, users may find themselves reaching a cap quickly, especially if they are attempting to generate long text summaries or create code projects. In our testing, reaching the limit resulted in a pop-up notification, which included a typographical error, indicating when the limit would reset—typically about three hours later.
If you have been skeptical about the Gemini models until now, it may be worth giving Gemini 2.5 Pro (Experimental) a chance. After a period of striving to keep pace with OpenAI, Google appears to be making significant strides with this latest release. Given the rapid rollout of the experimental model, it is likely that a condensed Flash model will be introduced soon for everyday tasks, making AI even more accessible for all users.
