Hey, I’ve been playing around with the Grok-3-mini-beta today — and honestly, it’s doing really well so far. Haven’t thrown anything super tough at it yet, but for my usual dev stuff, it nails it
1 Like
How does it compare to claude 3.7?
1 Like
@ericzakariasson
As per previous questions above regarding the Grok-3-mini reasoning setting, can you add something like so?:
grok-3-mini-beta-low
grok-3-mini-beta-high
They didn’t do it for o3-mini so I doubt they’ll do it here. But back then it took them 1 day to confirm the reasoning, while here I asked it 4 days ago.
Can @deanrie maybe answer?
Just gave a spin to gpt-4.1 and grok-3-beta - both miserably failed basic devops / repo restructuring task and deleted project files with no backup. Sonnet-3.7 did it with no issues and still made the backup
2 Likes
I just switched away from Sonnet 3.7 to o3 and it was an improvement, then I tried grok 3 mini, and found I liked it better than o3 (and faster)
Same experience here. I’d created a rule with rule type of “Agent Requested”. This type of rule is always available to the agent but the agent decides if it wants to use it based on the description. Grok-3-mini-beta picked up the rule while GPT-4.1 did not. Grok-3 models have a personality which makes them interesting to work with.
Will either models eventually support vision?