Google updates its Gemini AI model to improve capabilities for coding and creating interactive web apps
Google calls it the best coding model it has ever built.
It feels like every time you blink, there’s a new AI model claiming to be the smartest yet—and this week, it’s Google stepping up. Just ahead of its annual I/O conference, Google has dropped an updated version of its Gemini 2.5 Pro model, dubbed the “Gemini 2.5 Pro Preview I/O edition.” While that name might sound like just another label, this release is part of a deeper race that’s heating up across the AI space.
The updated Gemini 2.5 Pro Preview is more than a modest tweak; it comes with improved coding capabilities. Google calls it the best coding model it has ever built, with the model ranking number 1 in the WebDev Arena Leaderboard and LMArena in Coding.
Video credit: Google
This makes it very good at creating interactive web apps. Apart from coding, Gemini 2.5 Pro Preview also excels in video understanding, ranking at the top of the VideoMME benchmark with an 84.8%.
The new AI model is available through the Gemini API, Vertex AI, and AI Studio platforms. Interestingly, it costs the same as its predecessor, the Gemini 2.5 Pro. It can also be found in the Gemini chatbot across web and mobile apps.
As the AI race intensifies, competitors are also making significant strides. For example, OpenAI recently launched GPT-4.5, a new version of its AI model with significantly reduced hallucination rates, marking a substantial improvement over its predecessor. On the other hand, Meta released new Llama 4 AI models (Scout and Maverick) that push the limit to up to 17 billion parameters.
Louis Eriakha
However, everyone’s gunning for that same goal: an AI that doesn’t just respond but actually reasons through complex tasks like a human would (or better).
For Google, this isn’t just about releasing a shinier model; it’s part of a longer pivot. Less than two years ago, the company was playing catch-up after OpenAI’s ChatGPT made waves.
Now, it’s back in the lead—or at least trying to claim that position—by investing in core infrastructure and making smarter Gemini models. Meanwhile, for users, this means access to more powerful tools that can enhance productivity and creativity.
Oyinebiladou Omemu
