New Gemini models are quite good, Gemini 2.5 Pro is 1st in the user-benchmarks [1]. They also have Gemma, very good model that can run locally [2]. Benchmarks are not oracles of truth, but I feel like Google is not a kid who arrived late at the party anymore.
Yeah, with Gemini 2.5 Google stepped up to the grown up AI table and added on top. I still have a soft spot for Claude for general purpose chats, but have fully switched to Gemini for dev.
Considering that several key breakthroughs happened at Google that made GPT-style LLMs possible (e.g. Attention is all you need paper), it's more like they took a long smoke break than showed up late.
New Gemini models are quite good, Gemini 2.5 Pro is 1st in the user-benchmarks [1]. They also have Gemma, very good model that can run locally [2]. Benchmarks are not oracles of truth, but I feel like Google is not a kid who arrived late at the party anymore.
[1] https://lmarena.ai/
[2] https://blog.google/technology/developers/gemma-3/