cpatonn

/

Qwen3-30B-A3B-Thinking-2507-AWQ-4bit

like 13

Hey there. I'm struggling to get any thinking out of this model, it just behaves like an instruct. I'm running it in vllm with the following relevant flags:

      --model cpatonn/Qwen3-30B-A3B-Thinking-2507-AWQ-4bit
      --max-model-len 131072
      --gpu-memory-utilization 0.90
      --reasoning-parser qwen3
      --enable-auto-tool-choice
      --tool-call-parser hermes