1-bit quant of Hy3 295B runs 2.2x faster than cloud API without quality loss

Community quantization of Tencent's Hy3 295B model to 1-bit produces a 92GB IQ1_M GGUF file that runs locally on 4x RTX 5090. In tests, the quantized model matched the cloud API's quality on a retro game generation task while running 2.2x faster. The result suggests extreme quantization can preserve capability for some workloads.