跳到主内容
@wquguru
精选88MarkTechPost(RSS)产品发布/更新

OpenAI发布GPT-Live-1.1语音模型,引入待机计费模式

OpenAI Ships GPT-Live-1.1 for Voice Agents, Cuts Idle Cost to $0.02/Min

原文
发到 X
推荐理由

语音Agent开发者必看,这次引入了更便宜的待机计费模式,能显著降低长连接闲置成本,赶紧评估你的链路架构。

OpenAI has made GPT-Live-1.1 available in its API under the model ID gpt-live-1.1. The GPT-Live-1.1 voice model is a full-duplex model for multilingual conversations. It can listen while it speaks. OpenAI team states that it speaks with more natural and consistent accents across languages and completes tasks more reliably. The release also adds standby, a cheaper billing state for idle sessions. For teams running voice agents, idle time on a call now costs less. You can check the deployment options and specs here through our GPULeaderboard.com chrome extension.

OpenAI 已在其 API 中以模型 ID gpt-live-1.1 发布了 GPT-Live-1.1。GPT-Live-1.1 语音模型是一款用于多语言对话的全双工模型,能够在说话的同时进行聆听。OpenAI 团队表示,该模型在跨语言时能使用更自然且一致的口音,并能更可靠地完成任务。此次发布还新增了待机(standby)模式,这是一种针对空闲会话的更便宜的计费状态。对于运行语音代理的团队来说,通话中的空闲时间现在费用更低。你可以通过我们的 GPULeaderboard.com Chrome 扩展程序在此处查看部署选项和规格。

TL;DR

TL;DR(太长不看版)

  • Size: Live sessions hold a 128,000-token default context window.
  • Runs on: OpenAI’s cloud only, via v1/live/sessions. Clients connect over WebRTC or WebSocket. No open weights, no quantized variants.
  • Performance: OpenAI claims better accents, transcription, letter and number handling, and task completion than GPT-Live-1. No scores published.
  • Best: Standby bills idle WebRTC time at $0.02 per minute versus $0.05 active, a 60% lower rate.
  • Worst: OpenAI published 0 benchmark scores for 1.1. Image, video and structured outputs are not supported.
  • Bottom line: Best: cheaper always-on voice sessions. Worst: quality gains are self-reported with no numbers.
  • 规模:实时会话默认上下文窗口为 128,000 个 token。
  • 运行环境:仅限 OpenAI 云端,通过 v1/live/sessions 接口访问。客户端通过 WebRTC 或 WebSocket 连接。不提供开源权重,也无量化变体。
  • 性能:OpenAI 声称其在口音、转录、字母和数字处理以及任务完成方面优于 GPT-Live-1。未公布具体评分。
  • 最佳:待机模式对空闲 WebRTC 时间的计费为每分钟 0.02 美元,而活跃状态为每分钟 0.05 美元,费率降低了 60%。
  • 最差:OpenAI 未公布 1.1 版本的任何基准测试分数。不支持图像、视频和结构化输出。
  • 总结:最佳点是始终在线的语音会话成本更低;最差点是质量提升仅为自我报告,缺乏数据支持。

What is GPT-Live-1.1?

什么是 GPT-Live-1.1?

GPT-Live-1.1 is a full-duplex speech model from OpenAI that hears and speaks in 1 stream. It follows GPT-Live-1, which reached the API on September 10, 2026. As per the OpenAI forum announcement, it has powered ChatGPT Plus and Pro voice since October 15.

GPT-Live-1.1 是 OpenAI 推出的一款全双工语音模型,可在单一流中同时听和说。它接替了于 2026 年 9 月 10 日上线 API 的 GPT-Live-1。根据 OpenAI 论坛的公告,自 10 月 15 日起,它已为 ChatGPT Plus 和 Pro 版的语音功能提供支持。

Inputs and outputs are audio and text. Image and video are not supported. Streaming and function calling work. Structured outputs, fine-tuning and predicted outputs do not.

输入和输出均为音频和文本。不支持图像和视频。支持流式传输和功能调用。不支持结构化输出、微调及预测输出。

How does GPT-Live-1.1 work?

GPT-Live-1.1 是如何工作的?

The GPT-Live family replaces the usual cascade of speech-to-text, an LLM and text-to-speech. OpenAI says the model reasons over incoming and outgoing audio together in a single model. It also handles turn detection and returns ASR transcripts natively.

GPT-Live 系列取代了通常由语音转文本、大语言模型和文本转语音组成的级联架构。OpenAI 表示,该模型在一个模型内同时对输入和输出的音频进行推理。它还原生处理话轮检测并返回 ASR(自动语音识别)转录结果。

Deeper work is delegated. The voice model can hand reasoning and tool calls to a backend text model. OpenAI names GPT-6 Astra for complex issues and Luna for high-volume tasks. Third-party backends also work.

更深层次的推理工作被委托出去。语音模型可以将推理和工具调用委托给后端的文本模型。OpenAI 将复杂问题交由 GPT-6 Astra 处理,将高容量任务交由 Luna 处理。第三方后端同样适用。

The Live conversations guide sets delegation.type to client or responses. Responses mode uses OpenAI models. Client mode connects your own backend. Instructions can run up to 16,384 tokens.

Live 对话指南将 delegation.type 设置为 client 或 responses。Responses 模式使用 OpenAI 的模型;Client 模式则连接你自己的后端。指令最多可运行 16,384 个 token。

What changed from GPT-Live-1?

与 GPT-Live-1 相比有哪些变化?

OpenAI lists 4 improvement areas for 1.1:

OpenAI 列出了 1.1 版本的 4 个改进领域:

  • More natural and consistent accents across languages
  • More accurate speech transcription
  • Better handling of precise details like letters and numbers
  • More reliable instruction following and task completion
  • 跨语言口音更加自然且一致
  • 更准确的语音转录
  • 更好地处理字母和数字等精确细节
  • 更可靠地遵循指令并完成任务

How does standby mode work?

待机模式如何工作?

Standby pauses inference during silence while the WebRTC connection stays open. The guide documents 3 settings:

在 WebRTC 连接保持打开的状态下,待机模式会在静音期间暂停推理。指南中记录了 3 项设置:

  • session.wake_policy.type: set to speech, so inference waits for speech
  • session.idle_timeout_s: a whole number from 5 to 60 seconds
  • session.wake_policy.vad_threshold: optional, 0 to 1, default 0.5
  • session.wake_policy.type:设置为 speech,以便推理等待语音输入
  • session.idle_timeout_s:5 到 60 秒之间的整数
  • session.wake_policy.vad_threshold:可选参数,范围 0 到 1,默认值 0.5

After the timeout, the server emits session.paused. When the user speaks, session.resumed fires. The session ID and history are kept. OpenAI warns the first reply after a short utterance may be slower. Keep the microphone track active during standby.

超时后,服务器会发出 session.paused 事件。当用户开始说话时,触发 session.resumed 事件。会话 ID 和历史记录均被保留。OpenAI 警告称,短语音后的首次回复可能会较慢。请在待机期间保持麦克风轨道处于活动状态。

Standby works for both GPT-Live-1 and GPT-Live-1.1.

待机模式适用于 GPT-Live-1 和 GPT-Live-1.1。

How much does GPT-Live-1.1 cost?

GPT-Live-1.1 的费用是多少?

Session statePrice per minutePrice per hour
Active$0.05$3.00
Standby (WebRTC, inference paused)$0.02$1.20
会话状态每分钟价格每小时价格
活跃状态$0.05$3.00
待机状态(WebRTC,推理已暂停)$0.02$1.20

Voice sessions are billed by duration. The GPT-Live-1 model page says billing is per second. Backend model and tool usage are billed separately. That matters: a heavy backend can cost more than the voice layer.

语音会话按持续时间计费。GPT-Live-1 模型页面指出计费单位为每秒。后端模型和工具使用费用单独计算。这一点很重要:繁重的后端成本可能高于语音层。

Deployment and availability

部署与可用性

The model is API-only. You switch an existing Live session by setting model to gpt-live-1.1. Changing the model requires a new session.

该模型仅通过 API 提供。你可以通过将 model 设置为 gpt-live-1.1 来切换现有的 Live 会话。更改模型需要创建新会话。

Outbound SIP calling reached all organizations on November 2, so phone agents are supported. GPT-Live-1 and GPT-Live-1.1 have separate concurrent-session limits. Each limit applies across inference regions. The 1.1 page does not publish tier numbers.

11 月 2 日,出站 SIP 呼叫已覆盖所有组织,因此支持电话代理。GPT-Live-1 和 GPT-Live-1.1 拥有独立的并发会话限制。每个限制均跨推理区域适用。1.1 页面未公布层级数量。

How does GPT-Live-1.1 compare?

GPT-Live-1.1 有何不同?

FeatureGPT-Live-1.1GPT-Live-1GPT-Realtime-2.1Gemini 3.8 Live
DeveloperOpenAIOpenAIOpenAIGoogle
ParametersNot disclosedNot disclosedNot disclosedNot disclosed
InputAudio, textAudio, textText, audio, imageText, image, audio
OutputAudio, textAudio, textText, audioText, audio
Context128K default (guide)128K default (guide)128K, 32K max output131,072 in, 65,536 out
Pricing$0.05/min active, $0.02/min standby$0.05/min active, $0.02/min standby$32 / $64 per 1M audio tokens (in/out)Token-based (pricing)
Idle standby billingYes (WebRTC)Yes (WebRTC)Not documentedNot documented
ReasoningDelegated to backend modelDelegated to backend modelBuilt in, configurable effortInterleaved reasoning
Function callingYesYesYesYes
Structured outputsNoNoNot disclosedNo
Published benchmarksNone for 1.1Full Duplex Bench +30 pts vs GPT-Realtime-2.1Baseline in that evalNot compared
License / weightsClosed, API onlyClosed, API onlyClosed, API onlyClosed, API only
功能GPT-Live-1.1GPT-Live-1GPT-Realtime-2.1Gemini 3.8 Live
开发者OpenAIOpenAIOpenAIGoogle
参数未公开未公开未公开未公开
输入音频、文本音频、文本文本、音频、图像文本、图像、音频
输出音频、文本音频、文本文本、音频文本、音频
上下文窗口默认 128K(指南)默认 128K(指南)128K,最大输出 32K输入 131,072,输出 65,536
定价活跃状态 $0.05/分钟,待机状态 $0.02/分钟活跃状态 $0.05/分钟,待机状态 $0.02/分钟每百万音频 token(输入/输出)$32 / $64基于 token 计费(定价)
空闲待机计费是(WebRTC)是(WebRTC)未记录未记录
推理能力委托给后端模型委托给后端模型内置,可配置推理强度交错式推理
函数调用支持支持支持支持
结构化输出不支持不支持未公开不支持
已发布的基准测试1.1 版本无Full Duplex Bench 比 GPT-Realtime-2.1 高出 30 分该评估中的基线水平未进行比较
许可证/权重闭源,仅限 API闭源,仅限 API闭源,仅限 API闭源,仅限 API

更进一步:量化金融体系

看懂新闻只是起点——沿量化金融路径,把它变成能交付的工程能力

进入量化体系 →

相似阅读

关联信息,但可能不是同一事件