akitaonrails starred syv-ai/HyperQwen · October 28, 2026 08:27 syv-ai/HyperQwen Serve large Qwen models fast on the GPUs you actually own. Qwen3.8-27B on a single 24 GB card with vLLM: 127 tok/s single-user (381 when the answer… Python 2.4k Updated Oct 21