hey yall. So I have a 24GB gpu. What do you think is better? I am using unsloth quants. Both are UD quants. I need 262K context for my hermes agent and use case. Both setups fit perfectly in vram.
I have heard that Qwen 3.6 27B is quite good even with Q4 KV.
I am using LM studio so I need need to use V and K at the same value or else CPU usage goes much higher.
[link] [comments]