Event ·Single source

fractalyze_io 优化 Qwen3-Omni,单卡 RTX 5090 首音频延迟降至 23ms

fractalyze_io 在 vLLM-Omni 上对 Qwen3-Omni 做了优化,可在单张 RTX 5090 上运行。他们采用 AWQ-4bit 量化,在 batch-1 文本提示测试中,把首个音频的响应时间从原版 vLLM-Omni 的 213ms 降到 23ms。vLLM 团队呼吁将这些优化贡献回上游 vLLM-Omni。

Current read

有人在单张 RTX 5090 上跑 Qwen3-Omni,用 AWQ-4bit 把首音频延迟从 213ms 压到 23ms,本地语音对话流畅度差了近十倍。

2 sources42° AI heatLast updated 2026/10/5 13:27:27

Evidence

2 sources
Related source 1@atomic_chat_hq
View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.