Google introduces Gemini 2.5 Pro, and it created a game in a minute

Google introduced Gemini well over a year ago, and this AI platform has been getting smarter as time went on. There’s no doubt that Google’s AI is some of the smartest on the market, even though the company has had some massive missteps. On Tuesday, Google announced the new Gemini 2.5 Pro model, and this will power new AI experiences going forward.

This model’s full name is Gemini 2.5 Pro Experimental. As such, this is an experimental model. While it will be smarter than Gemini 2.0, you should still expect some minor issues to pop up. AI isn’t perfect, and it looks like Google is testing the waters with this new model.

Google introduced Gemini 2.5 Pro Experimental

At this point, we don’t know exactly how capable this model is, but Google touts that it sits on top of the LMArena leaderboard by a “significant margin.” This means that there’s definitely some brains behind this model. It’s the company’s advanced model for completing complex tasks, and this is something that all AI companies are working on. Not only that, but it topped benchmarks like GPQA and AIME 2025.

One area where it excels is in the coding area. Gemini has always had a focus on coding, but version 2.5 Experimental takes it to another level. The company posted a video showing Gemini coding a simple infinite runner purely from a text prompt. The video is posted below:

Gemini made the game in one minute and seven seconds. While the infinite runner was infinitely simple, it required nothing more than a short text prompt. Looking at the cursor in the video, we can tell that it was an animation, not a screen recording, so you’ll want to take it with a grain of salt.

You might not be able to generate your MMORPG with Gemini, but it shows how far the tech has come. Just like the Gemini models that came before, Gemini 2.5 Pro Experimental has a robust multimodality with the ability to understand images, text, audio, video, and code repositories. Not only that, but users will still have access to the massive 1 million-token context window.

Cost/availability

If you want to get your hands on this model, there are a few things to know. Firstly, it’s not available to free users. Gemini Advanced users can use the model in the Gemini app and Google AI Studio. It will expand to Vertex AI soon.

Along with that, Google will reveal the pricing for other ways to access this model soon. At this point, we don’t know if it will be expensive or affordable.

The post Google introduces Gemini 2.5 Pro, and it created a game in a minute appeared first on Android Headlines.