Google is pretty far behind. They have random one off demos and they beat benchmarks yes, but try to use Google’s AI stuff for real work and it falls apart really fast.
The only thing OpenAI has right now is the ChatGPT name, which has become THE word for modern LLMs among lay people.
Anecdotally, I've switched to Gemini as my daily driver for complex coding tasks. I prefer Claude's cleaner code, but it is less capable at difficult problems, and Anthropic's servers are unreliable.