返回首页

ChatGPT to Undergo Its Biggest-Ever Overhaul, Evolving From a Chatbot Into a Super App

ChatGPT to Undergo Its Biggest-Ever Overhaul, Evolving From a Chatbot Into a Super App

Source: Jiemian News
According to a report by CCTV Finance on June 7, OpenAI is set to roll out the most extensive upgrade to ChatGPT since its launch in 2022, with the new version to be released gradually in the coming weeks.
OpenAI plans to transform ChatGPT into a super app by integrating programming tools and AI agents, and launching a suite of new products to diversify revenue streams. A major focus of this upgrade is its coding product Codex, which will receive significantly increased resource allocation. This marks a strategic shift: OpenAI’s offerings will no longer be limited to conversational Q&A chatbots, but center on intelligent agents capable of autonomously completing a wide range of tasks for users.
Codex is OpenAI’s AI code generation model built on an enhanced GPT-3 architecture, specialized in converting natural language prompts into code across various programming languages. Its beta version was released via API in August 2021, followed by the open-source local tool Codex CLI in April 2025. The full official launch took place last October. To date, it has processed 40 trillion tokens, with added features including Slack integration, Codex SDK and management tools that support enterprise-level environment control, analysis and monitoring. It has since become one of OpenAI’s fastest-growing products and a major revenue driver.
While pressing ahead with IPO preparations, OpenAI faces mounting pressure to boost revenue and build sustainable profitability ahead of its public listing.
Citing reports from The Information, OpenAI generated approximately $5.7 billion in revenue in the first quarter of this year, nearly $1 billion more than Anthropic over the same period. Its growth was mainly fueled by Codex sales, enterprise client business and pilot advertising on ChatGPT.
Nevertheless, the company’s profitability remains bleak. Its adjusted operating profit margin stood at -122% in Q1. Even after excluding certain major costs, OpenAI lost around $1.22 for every dollar of revenue. Substantial spending on model training, inference computing power, business expansion and user subsidies continues to burn through capital.
By contrast, Anthropic is projected to post roughly $600 million in operating profit in the second quarter with an operating margin of about 5.5%, edging close to break-even.
OpenAI has also hit a ceiling in user growth. Its average weekly active users stood at around 905 million in Q1, down from a peak of 920 million recorded in February. On the paid side, ChatGPT had 55 million consumer subscribers in the first quarter, up from roughly 47 million at the end of last year.
A year ago, the industry race centered on building more powerful AI models. Today, the market focuses on two core goals: turning AI usage into revenue, and converting revenue into steady profits. This means leading AI firms must target high-value customer groups instead of relying solely on free chatbot users.
Anthropic serves as a typical example. To win over Wall Street clients, Anthropic PBC has rolled out new AI agents tailored for a broad spectrum of financial services tasks, such as drafting presentation materials for client meetings, reviewing financial statements and escalating cases for compliance reviews.
The company is optimizing the Claude model for seamless interoperability with third-party software including Excel, PowerPoint and Outlook, and integrating data from financial partners such as Dun & Bradstreet and Moody’s. It is also deepening its footprint in the financial sector via new joint ventures with Blackstone, Hellman & Friedman and Goldman Sachs to expand enterprise adoption of its software.
Google is likewise expanding the application of its AI agents across more scenarios. At Google I/O 2026 held in May, the tech giant announced that AI agents will be embedded throughout all core entry points — from search bars and the Chrome browser to Android devices and smart glasses. Gemini is evolving beyond a mere chat assistant into a persistent cross-app AI agent that tracks information, generates content, invokes tools and even completes orders and operational workflows on users’ behalf.
Google CEO Sundar Pichai stated at the developer conference that the past twelve months have been a period of rapid progress for the company. Google now processes 32 quadrillion tokens per month, a seven-fold year-on-year increase, while Gemini has reached 900 million monthly active users.
Just like Anthropic and Google, OpenAI is eager to tap new profit drivers. As its largest revamp to date, the company’s heavy bet on Codex remains to be seen whether it can deliver the expected commercial value.

ChatGPT将迎史上最大改版,从单一聊天变身“超级应用”
来源:界面新闻
据央视财经6月7日报道,OpenAI即将对旗下聊天机器人ChatGPT进行自2022年上线以来规模最大的一次升级,新版将在未来几周内逐步上线。
据悉,OpenAI计划将ChatGPT转型为“超级应用”,整合编程工具与人工智能体,新增多款产品,以拓宽营收渠道。此次升级将重点加码旗下编程产品Codex,加大对其资源倾斜力度。这意味着,OpenAI的产品不再局限于问答类聊天机器人,而是主打可自主帮用户完成各类任务的智能体。
Codex是OpenAI推出的AI代码生成训练模型,基于GPT-3架构改进,专注于将自然语言指令转换为多种编程语言代码。2021年8月,Codex以API形式发布测试版,2025年4月推出开源本地工具Codex CLI。去年10月,Codex正式版发布,并累计处理40T token,新增Slack集成、Codex SDK及管理工具,支持企业级环境控制与分析监控,成为OpenAI增长最快的产品之一。目前,该产品是OpenAI的主要收入来源之一。
OpenAI正在推进上市筹备工作。与此同时,该公司面临日益增长的压力,急需提升收入并开辟盈利道路,为首次公开募股做准备。
据The Information此前报道,OpenAI今年一季度收入约57亿美元,比Anthropic同期高出近10亿美元。OpenAI的增长主要来自Codex、企业客户销售,以及ChatGPT广告测试。
不过,OpenAI的盈利情况并不乐观。该公司一季度调整后经营利润率为-122%,即剔除部分大项成本,OpenAI每收入1美元,仍亏损约1.22美元。AI模型训练、推理算力、产品扩张和用户补贴,仍在吞掉大量现金。与之相比,Anthropic二季度可能实现约6亿美元经营利润,利润率大约5.5%,已经接近盈亏平衡线。
此外,OpenAI用户增长也陷入瓶颈。OpenAI一季度平均每周活跃用户约9.05亿,2月曾达到约9.2亿。付费用户方面,ChatGPT一季度有5500万名消费者订阅用户,高于去年底约4700万。
如果说,一年前,AI模型的核心叙事是谁能把模型做得更强,那么今年市场更关注的则是,谁能把AI使用量变成收入,谁能把收入变成更稳定的利润。这就意味着AI巨头们必须寻找能够创造高利润的用户,而不能局限于免费的聊天客户群。
以Anthropic为例。据此前报道,为争取华尔街客户,Anthropic PBC发布一系列新的人工智能智能体,旨在处理更广泛的金融服务任务。该公司的AI智能体可用于撰写客户会议推介材料、审阅财务报表,以及将个案升级至合规审查。
Anthropic表示,将使 Claude模型更好地与Excel、PowerPoint和Outlook 等第三方软件协同工作,整合来自邓白氏和穆迪等金融行业合作伙伴的数据。Anthropic还在通过与黑石集团、Hellman & Friedman以及高盛集团成立新的合资企业,加深在金融行业的布局。该合资公司将推动其软件在更多企业中的应用。
谷歌也正在将AI智能体应用到更多场景。在今年5月举行的谷歌I/O 2026开发者大会上,谷歌宣布正把AI智能体全面“塞进”所有核心入口。从搜索框到Chrome浏览器,从Android手机到智能眼镜,Gemini不再只是一个对话助手,而是一个可以持续运行、跨应用执行任务的AI代理:它能替用户追踪信息、生成内容、调用工具,甚至直接完成下单和操作流程。
谷歌CEO桑达尔·皮查伊(Sundar Pichai)在上述开发者大会上表示,过去12个月是谷歌“飞速发展”的一年。谷歌每月处理的Token数量达到3.2千万亿个,同比增长7倍;Gemini应用月活跃用户数达9亿。
和Anthropic、谷歌一样,OpenAI也需要寻找到新的利润增长点。作为史上最大改版,此次其对Codex加大押注,能否创造出预期中的价值,有待继续观察。