Gemini 2.5 Pro released

post by Yair Halberstadt (yair-halberstadt) · 2025-03-26T07:56:44.964Z · LW · GW · 0 comments

This is a link post for https://blog.google/technology/google-deepmind/gemini-model-thinking-updates-march-2025/

Contents

No comments

Google has now experimentally released Gemini 2.5 Pro.

As is standard, they claim it #beats all the benchmarks.

In my initial testing it's a very good model, definitely on par with GPT 4.5. It's a reasoning model, but runs very quickly, and shows you the chain of thought, unlike OpenAI models. I would consider using it for my daily driver. Biggest issue is that it does extensive thinking even when not really necessary (but as discussed it's quick so that matters less).

See more details at https://blog.google/technology/google-deepmind/gemini-model-thinking-updates-march-2025/.

Pricing is no yet available.

Detailed table displays performance of multiple large language models on tests like math, coding, and reasoning. Gemini 2.5 Pro shows top results in several categories, indicated by highlighted cells. Fine print at the bottom provides context for the data.

0 comments

Comments sorted by top scores.