AI News AI资讯 1d ago Updated 23h ago 更新于 23小时前 49

OpenAI wants its new tool to do your work for you and with you OpenAI希望新工具为你和你一起完成工作

OpenAI launches ChatGPT Work, an autonomous agent capable of sustaining long-running tasks for hours, addressing previous limitations in its Atlas browser's "Agent Mode." The platform integrates scheduled tasks, deep connectivity with workplace tools (Slack, Teams, Drive), and a new GPT-5.6 model tiered for performance and cost efficiency. OpenAI is restructuring its ecosystem by merging Codex into ChatGPT Work, sunsetting the standalone Atlas browser, and rebranding the legacy desktop app as "C OpenAI发布ChatGPT Work,解决此前Agent模式任务中断问题,支持长时间运行复杂项目并自动完成工作流。 集成Scheduled Tasks功能,支持定时或事件触发的重复性任务自动化,并可跨设备监控,同时整合Slack、Teams等常用办公工具插件。 推出新模型GPT-5.6,提供更高性价比的性能,采用三级定价策略,最高档输入/输出价格分别为$5/$30每百万token。 原有桌面版ChatGPT更名为“Classic”,编码能力Codex合并入Work视图,复杂工作任务将消耗更多订阅额度,企业版支持设置支出上限。

75
Hot 热度
65
Quality 质量
70
Impact 影响力

Analysis 深度分析

TL;DR

  • OpenAI launches ChatGPT Work, an autonomous agent capable of sustaining long-running tasks for hours, addressing previous limitations in its Atlas browser's "Agent Mode."
  • The platform integrates scheduled tasks, deep connectivity with workplace tools (Slack, Teams, Drive), and a new GPT-5.6 model tiered for performance and cost efficiency.
  • OpenAI is restructuring its ecosystem by merging Codex into ChatGPT Work, sunsetting the standalone Atlas browser, and rebranding the legacy desktop app as "ChatGPT Classic."
  • Usage is monetized via a credit-based subscription model up to $100/month, with enterprise controls allowing granular spending limits to manage high token consumption.

Why It Matters

This release marks a strategic pivot for OpenAI from conversational assistants to persistent, workflow-integrated agents, signaling a shift toward automating complex, multi-step professional tasks rather than simple queries. For AI practitioners, it highlights the growing importance of context persistence, tool integration, and cost management in production-grade AI applications. The introduction of GPT-5.6 and specific pricing tiers demonstrates OpenAI's effort to balance high-performance computing costs with enterprise adoption requirements.

Technical Details

  • Long-Running Agents: ChatGPT Work supports tasks that persist for hours, utilizing "Scheduled Tasks" similar to enhanced cron jobs that trigger on schedules or monitored events, allowing background processing accessible via mobile.
  • Tool Integration & Plugins: The system connects to Slack, Microsoft Teams, Google Drive, and SharePoint via custom plugins. Users can force specific tool usage with "@" mentions, while the desktop version accesses local files and a built-in browser.
  • Model Architecture: Launched alongside GPT-5.6, which offers three performance tiers. The highest tier costs $5 per million input tokens and $30 per million output tokens, optimized for "harder work."
  • Access Control: Implements a Compliance API and enterprise/admin controls to restrict access to external apps and files, ensuring security boundaries within autonomous workflows.
  • Ecosystem Restructuring: Codex technology is built into ChatGPT Work, while the old desktop interface is deprecated to "ChatGPT Classic," requiring explicit activation for basic chat functions.

Industry Insight

  • Cost Management is Critical: As agents run longer and consume more tokens, enterprises must implement strict budget controls. OpenAI’s allowance for group-level spend limits suggests that financial governance will be a primary concern for adopting autonomous agents.
  • Shift from Chat to Workflow Automation: The distinction between "Chat" and "Work" views indicates that future AI interfaces will be mode-dependent. Professionals should expect specialized environments for deep work versus casual interaction, impacting how teams train and deploy AI tools.
  • Consolidation of Developer Tools: Merging Codex into the main ChatGPT app simplifies the developer experience but may blur lines between general-purpose assistance and specialized coding tasks. Teams should evaluate whether the integrated environment meets their specific coding workflow needs compared to standalone solutions.

TL;DR

  • OpenAI发布ChatGPT Work,解决此前Agent模式任务中断问题,支持长时间运行复杂项目并自动完成工作流。
  • 集成Scheduled Tasks功能,支持定时或事件触发的重复性任务自动化,并可跨设备监控,同时整合Slack、Teams等常用办公工具插件。
  • 推出新模型GPT-5.6,提供更高性价比的性能,采用三级定价策略,最高档输入/输出价格分别为$5/$30每百万token。
  • 原有桌面版ChatGPT更名为“Classic”,编码能力Codex合并入Work视图,复杂工作任务将消耗更多订阅额度,企业版支持设置支出上限。

为什么值得看

本文标志着OpenAI从单纯的对话助手向长期自主代理(Autonomous Agent)平台转型的关键一步,解决了AI在长周期任务中的稳定性痛点。对于企业用户而言,理解其新的计费逻辑、权限控制机制及模型分层策略,是评估AI落地成本和合规风险的核心依据。

技术解析

  • 长时任务与自动化架构:ChatGPT Work旨在处理需数小时完成的复杂项目,通过“Scheduled Tasks”实现类似cron jobs的定时或事件驱动自动化,支持后台持续运行及移动端监控,弥补了早期Agent模式易中断的缺陷。
  • 生态集成与插件系统:工具通过自定义插件深度集成Slack、Microsoft Teams、Google Drive、SharePoint等主流办公软件,桌面版可直接访问本地文件,Chrome扩展支持浏览器内原生操作,并通过“@”符号强制调用特定第三方应用。
  • 模型性能与定价层级:伴随发布的GPT-5.6模型强调“每美元性能更强”,分为三个层级,最高性能层级定价为输入$5/百万token、输出$30/百万token,针对高难度工作优化成本效益。
  • 权限管理与成本控制:提供Compliance API及企业/管理员控制面板,允许限制对外部应用和文件的访问;企业版支持设置总体、群组或个人级别的支出上限,以应对长时间运行带来的高额Token消耗。

行业启示

  • AI Agent商业化路径清晰化:OpenAI通过区分“聊天”与“工作”视图及独立的计费体系,明确了AI从辅助工具向生产力引擎转变的商业逻辑,企业需重新评估AI使用的预算模型。
  • 企业级合规与安全成为标配:内置的权限控制和支出限制功能表明,AI在深入企业核心工作流时,数据安全和成本可控性是首要考量,IT治理策略需同步升级。
  • 模型迭代聚焦ROI而非单纯参数:GPT-5.6强调“每美元性能”,反映出行业竞争焦点正从追求极致能力转向平衡性能与成本,高性价比将成为企业大规模部署AI代理的关键驱动力。

Disclaimer: The above content is generated by AI and is for reference only. 免责声明:以上内容由 AI 生成,仅供参考。

GPT GPT Agent Agent Product Launch 产品发布