OpenAI 新智能体剑指 Meta，但能与免费模式竞争吗？

在 OpenAI 年度 DevDay 大会上，该公司全力应对竞争对手，主要是 Meta——其 Muse AI 智能体平台早期已取得爆发式成功。CEO Sam Altman 在欢呼声中走上舞台，发布了 Dots：一款由 GPT-6 Astra 驱动的“真正的 AI”智能体，其灵感来自“我们成长过程中在电影里看到的那些酷炫智能体”。与 Muse 类似，Dots 也试图以令人放下戒心的可爱设计吸引用户：它呈现为带眼睛、色彩鲜艳且可个性化定制的团块。OpenAI 通过演示 Dots 不仅能担任有用的助手、还能构建类似元宇宙的虚拟世界，向 Meta 发起挑战。但在一个关键领域，Dots 显然无法竞争：价格。

Muse 的核心卖点是免费且无摩擦地提供 AI 智能体访问服务。该公司提供免费的专用虚拟机，用于容纳智能体以及用户相关数据。相比之下，OpenAI 的 Dots 目前仅限每月 100 美元的 Pro 计划及更高等级用户使用。只要拥有 Meta 账户，任何人都可以使用 Muse；而 ChatGPT 的 12 亿用户中，大多数将无法使用 Dots。

原因很明显：OpenAI 很可能负担不起。根据 Bloomberg 的报道，这家 AI 实验室目前正寻求以 1.4 万亿美元估值筹集 300 亿美元资金；Axios 报道称，自第三季度初以来，该公司的增长率已达 70%，年度经常性收入接近 700 亿美元。然而，尽管融资规模惊人，其资金并非无限，尤其是与 Meta 这样的科技巨头相比。在过去一年里，一些用户指责 AI 实验室在 AI 资金紧缩 期间将成本转嫁给客户。OpenAI 尤其正在快速消耗大量现金——根据 FT，该公司预计到 2030 年将支出 2800 亿美元——而算力成本高昂，同时其在上市前还面临投资者要求改善盈利状况的压力。

“Dots 起步时会是一款高端产品——它消耗大量算力。”Altman 周二在 DevDay 主题演讲后的一场私下媒体问答中表示。他补充说：“我认为它将帮助我们理解，也帮助人们理解，AI 能做的事情有多广泛。但你们当然应该期待，有一天我们会为数十亿人推出面向大众市场的产品。”

在 DevDay 期间，OpenAI 着力解释 Dots（单个智能体称为一个“Dot”）为何物有所值，将其定位为不仅是一名能力出众的个人助理，更是一名“首席幕僚”或资深工程师。OpenAI 员工 Holly Li 告诉现场观众，使用 Dot“就像在和第二个大脑交谈”。Altman 表示，Dots 是为委托处理“有雄心的大块工作”而构建的，例如未发布产品的 QA、网站设计、团队异地活动行程规划、投资组合再平衡以及电子邮件营销活动创建。针对会计、法律分析等领域训练的“专家型”Dots 预计很快上线，公司还正与 Microsoft 合作，将它们整合进其 Agent 365 平台。

Altman 表示，其目标是超越那些只能预订服务、购买餐食、查询日历和安排预约的智能体——这些是 Muse 等面向消费者工具的典型应用场景。

在媒体问答中，当被问及 Dots 与 Muse 以及另一款热门新 AI 智能体 Instinct 有何不同时，OpenAI 核心产品负责人 Thibault Sottiaux 表示，Dots“实际上能够在非常长的时间跨度内保持任务不偏离轨道”。他补充称，Dots 的多模态能力将允许用户通过文本、语音模式和电话与智能体交流，这会让人感觉它是“你不可思议的延伸”。（不过，Dots 的语音演示在 DevDay 上未能成功运行；此外，尽管 OpenAI 高管在大会期间多次强调向智能体发短信的功能，这项已获 Muse 和 Instinct 支持的功能，对 Dots 而言仍计划“即将推出”。）

Altman 在问答中表示，长周期任务和知识工作任务正是 Dots 的差异化所在，并补充说：“我认为，这远远超过了当前市场上的任何其他产品。”他说，智能体在梳理待办事项和支持员工工作方面的帮助，“经济学家可能难以立即衡量”，但有望极大改变人们对自身工作的感受。

将 Dots 定位为专业工具，不仅使它区别于 Muse，也可能帮助 OpenAI 在企业市场对抗 Anthropic。Anthropic 多年来一直专注于企业 AI，并在这一领域占据主导地位；但在过去一年里，OpenAI 已向该领域投入更多资源。OpenAI CFO Sarah Friar 告诉 CNBC，自 7 月以来，该公司的企业业务已经翻倍。

OpenAI 高管还试图将 Dots 宣传为更安全、更具保障的产品，特别是相对于 Muse 而言。该公司 Codex 产品负责人 Alexander Embiricos 表示，OpenAI 的重点是打造“最值得信赖、最安全、最有保障的助手”，并强调 Dots 由该公司“对齐程度最高的模型”GPT-6 Astra 驱动。在 DevDay 主题演讲中，高管还展示了“OpenAI Private Intelligence”，这是一套数据隐私框架，可让企业对自身数据拥有“更强的控制权”，并提供零数据保留政策选项，即数据不会存储在 OpenAI 的服务器上。

Altman 在台上表示，OpenAI 希望“为前沿 AI 的隐私设立新标准”。此前，Muse 因接管一名用户的 Facebook Marketplace 账户、并在用户未察觉的情况下将其住址告知陌生人而受到强烈批评。

“我认为，我们与 Meta 所处的位置不同，因为他们没有一款拥有 12 亿用户的 AI 产品。”OpenAI 应用平台负责人 Glen Coates 在接受 The Verge 采访时表示，“推出会犯这类错误的产品，是我们会尽力谨慎避免的事情。我认为，我们正非常审慎地决定如何让智能体使用插件、如何让智能体使用你的身份信息并为事项登录或注册，以及如何让智能体替你付款。”

问题在于，这一切是否足以战胜免费。ChatGPT 的崛起不仅因为它抢得先机，也因为它免费——即使不登录也很容易使用。Muse 或许也能出于同样原因主导 AI 智能体领域。让这些 AI 智能体尽可能无摩擦的另一道障碍是，Meta 能够轻松整合其旗下其他平台，例如 Facebook、Instagram 和 WhatsApp。（Google 的 Gemini Spark 智能体平台也因同样的原因具备优势。）尽管 OpenAI 表示，用户将能够把 Dots 连接至广泛的外部工具，但这些连接未必能像内部整合那样无缝。

目前，OpenAI 希望更好的产品能够胜过更便宜的产品。Coates 表示：“我们的推进速度可能会稍微慢一点，但我们也知道，如果用户的 Dots 跑到 Facebook Marketplace 上随意购买东西，那绝不是我们想让 12 亿用户经历的事。基本上，我们是在努力把事情做对。”

**更正：**ChatGPT Pro 的价格为每月 100 美元，而非 20 美元。

At OpenAI’s annual DevDay conference, the company pulled out all the stops to compete with its rivals — primarily Meta, whose Muse AI agent platform has seen early runaway success. CEO Sam Altman walked onstage to cheers and announced Dots, a “real-deal AI” agent powered by GPT-6 Astra, “inspired by the cool agents that we all watched in movies growing up.” Similar to Muse, Dots aims for disarming cuteness, taking the form of colorful, personalizable blobs with eyes. OpenAI fired shots at Meta by demonstrating Dots’ ability to not just act as helpful assistants but build metaverse-esque virtual worlds. But in one key area, Dots strikingly fails to compete: price.

Muse’s main selling point is making AI agent access free and frictionless. The company offers a free dedicated virtual machine that houses both the agent and the user’s associated data. OpenAI’s Dots, by contrast, is restricted for now to users on its $100/month Pro plan and up. While anyone with a Meta account can get Muse, most of ChatGPT’s 1.2 billion users won’t be able to use Dots.

There’s an obvious reason for this: OpenAI likely can’t afford it. The AI lab is reportedly currently aiming to raise $30 billion in funding at a $1.4 trillion valuation, per Bloomberg, and it’s experienced 70 percent growth since the beginning of its third quarter — nearing $70 billion in annual recurring revenue, Axios reported. Yet despite its staggering financing, its funds aren’t limitless, particularly compared to a tech giant like Meta. In the past year, some users have accused AI labs of passing costs on to customers amid an AI money squeeze. OpenAI in particular is burning through a lot of cash right now — it expects to spend $280 billion by 2030, per the FT — and compute is expensive, while it’s also facing pressure from investors to button up its bottom line ahead of going public.

“Dots are starting out as a premium product — it uses a lot of compute,” Altman said Tuesday during a private press Q&A after his DevDay keynote, adding, “I think it will help us understand and help people understand how expansive and what AI can do. But you should of course expect us to do a mass-market thing for billions of people someday.”

During DevDay, OpenAI took pains to explain why Dots (with singular agents referred to as a “Dot”) are worth the money, positioning it as not just a capable personal assistant but a “chief of staff” or experienced engineer. OpenAI employee Holly Li told the audience using a Dot was “like talking to a second brain.” Altman said that Dots were built for delegating “ambitious pieces of work,” like QA for an unreleased product, website design, team offsite trip planning, portfolio rebalancing, and email campaign creation. “Specialist” Dots, trained for areas like accounting and legal analysis, are supposed to launch soon, and the company is working with Microsoft to integrate them into its Agent 365 platform.

The goal, Altman said, is to go beyond agents that simply book reservations, purchase meals, search calendars, and schedule appointments — classic use cases for consumer-facing tools like Muse.

When asked during the press Q&A how Dots is different from Muse and Instinct, another popular new AI agent, Thibault Sottiaux, OpenAI’s head of core products, said it’s “capable of actually staying on track in very long periods of time.” He added that the multimodality of Dots, which will allow chatting with an agent via text or voice mode and phone call, will feel like “an incredible extension of you.” (The voice demos of Dots didn’t work at DevDay, though, and though OpenAI leaders highlighted the texting-your-agent feature multiple times at the conference, that feature — already supported by Muse and Instinct — is still slated as coming soon.)

Altman said during the Q&A that the long-horizon and knowledge-work tasks were what set Dots apart, adding, “I think this is far beyond anything else available on the market today.” He said that the agents’ help with triaging to-do lists and supporting employees at work would be “difficult for economists to measure right away” but would hopefully make a “huge difference” in how people feel about their jobs.

Framing Dots as a professional tool not only differentiates Dots from Muse, but could help OpenAI gain ground against Anthropic in the enterprise market. Anthropic has focused on — and dominated in — enterprise AI for years, but OpenAI has devoted more resources to the area over the past year. OpenAI CFO Sarah Friar told CNBC that the company’s enterprise business had doubled since July.

OpenAI leads also tried to promote Dots as safer and more secure, particularly compared to Muse. Alexander Embiricos, the company’s Codex product lead, said OpenAI is focused on having the “most trustworthy, safe, and secure assistant,” highlighting that it’s powered by GPT-6 Astra, the company’s “most aligned model.” During the DevDay keynote, executives also presented “OpenAI Private Intelligence,” a data privacy framework that allows businesses to have “stronger controls” over their data, as well as zero data retention policy options, with no data stored on OpenAI servers.

Altman said onstage that OpenAI hopes to “set a new standard for privacy in frontier AI.” The comments come after Muse was recently lambasted for taking over a user’s Facebook Marketplace account and giving out their address to a stranger without the user realizing.

“I think we’re in a different position to Meta in that they don’t have an AI product that has 1.2 billion users,” Glen Coates, OpenAI’s head of app platform, told The Verge in an interview. “Launching something that makes those kinds of mistakes is something that we would try to take the care to avoid. I think we’re trying to be really deliberate with how we let our agents use plug-ins, how we let agents use your identity and sign-in/sign up for things, how we let your agents pay for things.”

The question is whether all that is enough to beat free. ChatGPT rose to prominence not only by being first but also by being free — and easy to use even without signing in. Muse may be able to dominate in the AI agent space for the same reason. The other hurdle for making these AI agents as frictionless as possible is that Meta offers easy integration with the other platforms it owns, such as Facebook, Instagram and WhatsApp. (Google’s Gemini Spark agent platform has an advantage for the same reason.) Though OpenAI said users would be able to connect Dots to a wide range of external tools, they may not be as seamless as an in-house integration.

For now, OpenAI is hoping that a better product will trump a cheaper one. “We might move a tiny bit slower, but we also know that if people have Dots that go out there and YOLO-buy stuff on Facebook Marketplace, that’s just not something we want to put 1.2 billion users through,” Coates said. “We’re trying to do it right, basically.”

Correction: ChatGPT Pro costs $100 per month, not $20.

Follow topics and authors from this story to see more like this in your personalized homepage feed and to receive email updates.

• Hayden Field

• AI

• Meta

• OpenAI

• Report

• Tech