Qwen3 0.6B (llama.cpp, CPU)

GGUF (Q4_K_S) via llama.cpp. OpenAI-ish endpoints at /v1/chat/completions and /v1/models.