Is there really a difference in coding performance between Claude 3.5 and GPT-4o?

Lately, I've been testing a bunch of things at work to add code review automation, and Claude definitely seems to do a good job of maintaining context across long files. But oddly, when it comes to solving algorithm problems, GPT feels like it writes cleaner code?

If anyone uses both, I'm curious which one you use as your main tool in practice.