I've only tried Grok Code Fast 1, so I can't speak for any of the other models.
In my experience, Grok is very fast and very cheap, but only moderately intelligent. It isn't stupid, but it rarely does anything that impresses me. The reason it's a useful model is that it is very, very fast (~90 tokens per second) and is very competitively priced.
conception
You should try cerebras with qwen. 2000 tokens/sec. It’s like chatting with the future usually- just an instant response.
In my experience, Grok is very fast and very cheap, but only moderately intelligent. It isn't stupid, but it rarely does anything that impresses me. The reason it's a useful model is that it is very, very fast (~90 tokens per second) and is very competitively priced.