Backend: transformers · Model: Qwen/Qwen3-8B. Uses free local inference or the Space's shared GPU; no paid API is configured.