Event ·Official first-hand

MOSS-Live 六项全双工语音指标领先，多轮对话仍落后 GPT-Live-1

复旦 MOSS 团队旗下 OpenMOSS 发布全双工语音模型 MOSS-Live，在六项全双工语音评测指标上排名第一，其中在背景人声干扰下答错的错误率仅为 2%，抗噪能力是其核心优势。评测同时揭示了短板：在多轮任务（multi-turn tasks）上，MOSS-Live 不及 OpenAI 的 GPT-Live-1；在音频理解维度则落后于字节跳动 Seed 团队的 Doubao-Seed-Duplex。这意味着 MOSS-Live 在抢话、插话等全双工交互指标上占优，但在需要持续理解上下文的复杂对话场景仍有差距。全双工语音是当前对话模型竞争的关键方向，GPT-Live-1、Doubao-Seed-Duplex 与 MOSS-Live 分别代表美中两国的不同技术路线，分项指标的分化也为开发者按场景选型提供了直接参照。

Current read

想给应用加语音交互的开发者值得看，MOSS-Live 抢话指标最强、背景噪音答错率只有 2%，但多轮对话比不过 GPT-Live-1，音频理解输给 Doubao，选型前能省不少试错成本。

11 sources29° AI heatLast updated 2026/11/14 02:00:26

Evidence

11 sources

Primary sourcepandaily

2026/11/14 02:00:26View source

Related source 1Decoder

2026/11/12 13:48:19View source

Related source 2Logan Kilpatrick

2026/11/12 18:06:55View source

Related source 3宝玉

2026/11/12 18:22:21View source

Related source 4vLLM

2026/11/12 21:24:44View source

Related source 5Shubham Saboo

2026/11/13 06:44:37View source

Related source 6IT之家

2026/11/13 08:41:34View source

Related source 7elvis

2026/11/13 14:50:50View source

Related source 8掘金本周最热

2026/11/12 04:08:48View source

Related source 9arXiv: OpenAI

2026/11/12 07:09:08View source

Related source 10百度 AI Baidu

2026/11/12 09:21:25View 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.

Open the story