Gemini 2.5

google drops an update

Posted by on April 27, 2025 · 3 mins read

Google Gemini 2.5

Last month, Google released an update to their Gemini LLM. As a subscriber to Google’s One service for file storage, my family and I currently have access to Gemini Advanced. My kids have found Gemini quite useful for techincal school work such as mathematics, but so far prefer OpenAI’s ChatGPT for subjects such as language and history. However, with both, they regularly have to verify the responses to eliminate incorrect statements.

2.0 Flash

For most quick questions and answers, we use the default 2.0 Flash model. As the fastest model, it is helpful for providing a useful response, much like Google Search of past decades. In fact, now Google Search now leads it’s website results with an “AI Overview” meaning one rarely has to click into page links anymore.

The Flash model is a step better as you can carry out simple discussions with it:

Pretty good for quick questions, however it’s not the new 2.5 model.

2.5 Flash Thinking

The new Flash model offers an expandable display of it’s Thinking process:

This addition is nice to have, particularly when one is trying to validate the result.

2.5 Pro

A bit later on the Pro model was also released, initially in AI Studio, but then to Gemini Advanced. Initial reports hailed its coding abilities, rivalling or even exceeding the current leader: Anthropic Claude Sonnet 3.7. In my testing to small program to scrape a website and generate a Google Calendar from the event listings, Gemini gave a number of good approaches:

In particular, the Python program generated actually worked great, after I followed the env setup directions:

thought it required some work in VS Code first. I’ll be working more with Gemini Pro 2.5 in the coming weeks…

More in this series…

← Previous Post