Jackrong

/

Qwopus3.5-27B-v3

like 219

vllm serve --model /opt/model/Qwopus3.5-27B-v3/ -tp 8 --port 8000 --max-model-len 174762 --reasoning-parser qwen3 --enable-auto-tool-choice --tool-call-parser qwen3_coder --gpu_memory_utilization 0.8 --served-model-name qw3.5

@XiaoZaiyi
Give it a try
--tool-call-parser hermes

--tool-call-parser hermes

this works, but it makes too many mistakes in opencode, every 3rd tool call fails due to model not following the format

@XiaoZaiyi
Give it a try
--tool-call-parser hermes
this works, thanks

@XiaoZaiyi
Give it a try
--tool-call-parser hermes

Work for me, thank you so much.