low-cost-hardware

A user demonstrates running the Qwen 27b AI model quantized to Q3_K_M on two RX 580 GPUs, achieving 7.39 tokens per second using old DDR3 hardware for under $100.