Event ·Multi-source

Grok Voice 3.0 发布：延迟低于 1 秒，Artificial Analysis 语音榜得分 86.9

SpaceXAI 发布语音转语音模型 Grok Voice 3.0，在 Artificial Analysis 语音综合榜单排名第一，得分 86.9。该模型全双工基准拿到 97%，τ-Voice 得分 71.6%，与 GPT-Live-1（Astra）相当但延迟降低 40%。在企业级 EVA 基准测试中取得 74% 的最高精度。API 定价为每分钟 0.08 美元，与 Grok Voice 2.0 持平，grok-voice-latest 用户自动升级。该模型每天已在特斯拉车内支撑数百万次对话，星链客服约 70% 来电无需人工介入。

Current read

语音模型又卷起来了，这个延迟不到 1 秒，价格还没涨，做语音智能体的可以去看看 benchmark 数据。

11 sources64° AI heatLast updated 2026/10/14 23:35:09

Evidence

11 sources

Primary sourceIT之家

2026/10/14 23:35:09View source

Related source 1Artificial Analysis

2026/10/14 17:26:20View source

Related source 2@cb_doge

2026/10/14 17:48:30View source

Related source 3Blognone

2026/10/13 01:08:47View source

Related source 4OpenAI Blog

2026/10/14 12:00:00View source

Related source 5TestingCatalog

2026/10/15 06:06:41View source

Related source 6Bindu Reddy

2026/10/13 00:39:49View source

Related source 7AIGCLINK

2026/10/13 02:01:17View source

Related source 8lmarena.ai

2026/10/13 02:39:07View source

Related source 9AI Will

2026/10/13 03:00:49View source

Related source 10DeepLearning.AI

2026/10/13 16:00:00View 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