Qwen

/

Qwen3-VL-30B-A3B-Thinking

like 131

Follow

Qwen 54.2k

I cannot get tag as well. I had the same issue with Qwen3-30B-A3B-Thinking before and it was simple to fix to add a tag to the chat_template.jinja. But it seems that for this model, fixing the chat_template.json doesn't work.

add --reasoning-parser deepseek_r1 to the command and it will work fine
vllm serve "Qwen/Qwen3-VL-30B-A3B-Thinking" --served-model-name him --trust-remote-code --gpu-memory-utilization 0.95 --max_model_len 30000 --max_num_seq 10 --reasoning-parser deepseek_r1