Okay hold on guys, you won’t believe this.
I was struggling with an issue regarding embedding images and sending to Milvus
I actually went back and forth for 10 mints with Claude 4 Sonnet (Thinking and Non-Thinking both), Gemini 2.5, o3
All failed
I gave the same prompt to grok-4
One prompt, one attempt, fixed !
Elon is not coming slow.
Yes, ít’ so powerful.
It is massively powerful, it’s fixed a couple scripts that previously were buggy messes with sonnet 4, I’m just hoping that its’ integration with Cursor gets smoother.
Does anyone know if Cursor’s allowing it to use the full context window, or is it clipped like some other models?
hope x-ai allows api access to the heavy version of grok 4 as well! and cant wait for their coding model, should be released in upcoming months!
And for me, it was the opposite. I spent 10 minutes going back and forth with Grok-4, and Sonnet solved it in one interaction.
That’s because grok family is not coding models, they will release a coding model in upcoming months.
anyone know the pricing of it yet?
lol
he can’t fix simple issues with Rust
I don’t know how to command him, he’s still a headless fly at the moment
It’s not a coding model. My problem specifically needed correct knowledge and a good web search which grok-4 seems to be great at. Generally, it’s able to do which no other model can. But for daily coding use, it’s not good.
lmao come on you’re trolling right. No one would use grok. You’d have to be an absolute lunatic to put your code through grok. right? This is a joke?
lmao it’s not good at all. it just wrote fake tests
Up to you if you wanna believe or not mate but this absolutely happened.
I used grok 4 for a frontend problem and it failed miserably. claude-4-sonnet is still too good for me
are you ready GAMERS?!? gRoK5 brought to you by raid shadow legends 
Screenshots