my only gripe rn is grok is still a shitty model to use. yeh it scores nearby openai and anthropic on benchmarks, but my personal experience has been underwhelming