I've been testing all day with new models to be able to save some money. I think we're all trying to do that.
Tried Kimi K 2.5
Tried GPT 5.4.
Tried Minimax 2.7
then i tried GLM 5.1(PRO)
And my goal is to not use it for coding but just as the execution and coordination agent that I talk back and forth with. My main agent.
All the other models seemed very generic, with no personality, and that bothered me.
But I finally found something today in GLM 5.1 and it's hard to even tell the difference between Sonnet and GLM 5.1.
So if you're on a budget and you're looking for a model that doesn't hallucinate and go crazy and forget who you are and break things, I say give GLM 5.1 a try as your main agent.
I paid for the $30/month plan, which has a 210,000 context token window.
It performs very well for my use case. again im not using it for coding. just coordination/execution.
hope this helps