I tried mlx engine, and the copied content is

xinference launch --model-name qwen2.5-instruct --model-type LLM --model-engine MLX --model-format mlx --model-size-in-billions 7 --quantization 4bit --n-gpu auto --replica 1 --request-limits  --gpu-idx

There is no value set for --gpu-idx and --requests-limit, for those with no value, don't add to command lines.

And the icons are too small maybe the users would never discover them, you can try to highlight them in some ways.

It should be --size-in-billions instead of --model-size-in-billions.

When I paste back the commands, the model size does not work.