I noticed Grok 4.1 it has higher precision and accuracy, so it is able to “get” what I’m asking, and give me the correct results. Whereas, grok-fast-4 wasn’t as good.
Grok-code seems better for writing specific changes. Better are writing code, but not good at planning and analysis.
I’d be a little surprised if the API providers are giving Cursor a slice. I think Cursor’s main incentive is to provide cheaper models so usage cost stays below the subscription fee. This doesn’t apply to Composer, of course. Though the primary incentive (keep costs below subscription fee) is still there. It’s just mixed in with taking some profit from for on-demand usage for Composer.
Probably because Grok is so cheap to use they won’t make any money on this. You will hit your subscription limit faster with other models and then run pay as you go or upgrade your plan. Seems to me they are just out for your money. We’ll see if they add Grok 4.2 when it is released in a week or two.
Meanwhile Grok Code Fast is free to use though. It’s alright model. But it’s not even as good as Cursors own model Composer 1.
What is going on here?
I can even set my custom AWS Bedrock API, but I cannot use xAI models except for the cheapest old model? This looks like a fig leave to me.
Where can I configure xAI models or at least the API? What is going on?