After using it a couple hours playing around, it is a very solid entry, and very competitive compared with the big US relaeses. I'd say it's better than GLM4.6 and I'm Kimi K2. Looking forward to v4
Did you try with 60k+ context? I found previous releases to be lacklustre which I tentatively attributed to the longer context, due to the model being trained on a lot of short context data.