ChatGPT, OpenAI's text-generating AI chatbot, has become a defining technology product of the decade. Since its public launch in November 2022, it has grown from a research curiosity into a service used by hundreds of millions of people every week. In 2025, ChatGPT moved from being a simple text tool to a multimodal assistant with voice, image generation, web search, shopping, and autonomous agent capabilities. The year also brought intense competitive pressure, legal disputes, and a major strategic pivot toward making ChatGPT the center of OpenAI's product universe.
At the start of 2025, ChatGPT had roughly 300 million weekly active users. By October, that number had climbed to 800 million. OpenAI also reported that more than one million businesses were using its products by November, making ChatGPT one of the fastest-growing enterprise platforms on record. The company's valuation and revenue expectations rose accordingly, with reports projecting revenue to triple to $12.7 billion in 2025.
January 2025: New models and early agents
The year began with the launch of o3-mini, OpenAI's latest reasoning model, designed to be both powerful and affordable. The company also introduced ChatGPT Gov, a version of ChatGPT Enterprise tailored for U.S. government agencies. This gave agencies easier access to OpenAI's tools while maintaining security, privacy, and compliance controls.
OpenAI also started testing phone-number-only signups and launched a beta feature called Tasks, allowing ChatGPT users to schedule reminders and recurring actions. Another notable addition was the ability to assign ChatGPT personality traits, such as 'chatty' or 'Gen Z.' Perhaps the most important release in January was Operator, a general-purpose AI agent that could take control of a web browser to book travel, make reservations, and shop online. The initial research preview was aimed at users on the $200-per-month Pro plan.
February 2025: Deep research and the end of o3
In February, OpenAI unveiled deep research, an agent designed to carry out complex, multi-step research tasks by analyzing information across numerous websites and sources. This was a major step toward making ChatGPT useful for professional analysts and knowledge workers. The company also allowed anyone to use ChatGPT web search without logging in, significantly reducing the barrier to access.
Perhaps more importantly, OpenAI canceled plans to release o3 as a standalone model. CEO Sam Altman said the company would instead focus on a 'unified' model called GPT-5, which would integrate the reasoning power of o3 into a single, simpler product offering. OpenAI also began revealing more of o3-mini's 'chain of thought' reasoning process, a response to competitive pressure from DeepSeek.
March 2025: Image generation goes viral
March was the month ChatGPT's image generation feature exploded in popularity. OpenAI rolled out a major upgrade powered by GPT-4o, allowing ChatGPT to generate and edit images directly in the chat. The feature went viral for its ability to create images in the style of Studio Ghibli and other iconic aesthetics, prompting a wave of AI-generated memes.
The success brought capacity challenges. Altman warned that the image generator's popularity would delay other product releases and cause occasional service slowdowns. OpenAI also relaxed some content moderation rules, allowing ChatGPT to generate images of public figures and certain sensitive subjects. At the same time, the company announced plans to release its first open-weight language model since GPT-2. New APIs for transcription and voice generation were also introduced, along with o1-pro, a more powerful and expensive reasoning model.
April 2025: GPT-4.1 and new reasoning models
OpenAI launched GPT-4.1, a family of models focused on coding and instruction following, with three sizes: GPT-4.1, mini, and nano. Although initially available only through the API, GPT-4.1 later became available in ChatGPT by popular demand. The company also released two new reasoning models, o3 and o4-mini, which could use ChatGPT tools like web browsing, image processing, and code execution. However, these models had higher hallucination rates than previous versions.
OpenAI added a 'library' section to ChatGPT for managing AI-generated images. The company also began rolling out memory features that let ChatGPT use information from previous conversations to personalize responses. In a more concerning development, OpenAI acknowledged a bug that allowed minors to generate graphic content with ChatGPT, and promised to fix it. The company also said it might adjust its safety standards if rivals released high-risk systems without comparable safeguards.
May 2025: Codex, GPT-4.1 in ChatGPT, and enterprise expansion
OpenAI introduced Codex, an AI coding agent powered by a new model called codex-1. Codex could complete software engineering tasks such as fixing bugs, writing features, and running tests, typically taking between one and 30 minutes. The release placed OpenAI in direct competition with tools like GitHub Copilot, Claude Code, and Cursor.
In May, OpenAI also brought GPT-4.1 and GPT-4.1 mini directly into ChatGPT in response to user requests. The company launched a data residency program in Asia for enterprise and educational customers, and announced OpenAI for Countries, a program to help governments build AI infrastructure and customize products for local needs. OpenAI also promised to address ChatGPT's growing sycophancy, following an update that made the chatbot overly flattering.
June 2025: Voice upgrades and o3-pro
June brought several upgrades to ChatGPT's voice mode. OpenAI enhanced Advanced Voice Mode for paid users, making conversations more natural and fluid. The company also added new features for business users, including connectors for Google Drive and Box, meeting recordings, and support for the Model Context Protocol. These features positioned ChatGPT as a more serious workplace assistant.
OpenAI also launched o3-pro, an enhanced version of its o3 reasoning model, for ChatGPT Pro and Team users. A notable infrastructure story emerged: OpenAI began using Google's AI chips to power ChatGPT, marking the first time the company relied significantly on non-Nvidia hardware. Research also raised questions about ChatGPT's effect on critical thinking, with an MIT study suggesting that users showed less cognitive engagement when writing with AI assistance.
July 2025: Agents, study mode, and open model delays
OpenAI introduced ChatGPT Agent, a general-purpose agent that could automate a wide range of computer tasks, including managing calendars, drafting presentations, running code, and shopping online. The company also unveiled Study Mode, a feature designed to encourage critical thinking by prompting students to engage with material instead of simply receiving answers.
However, the release of OpenAI's open-weight model was delayed again. Altman said additional safety testing was needed, and the launch was postponed indefinitely. Reports also emerged that OpenAI was planning to release an AI-powered web browser to compete with Google Chrome. By the end of July, ChatGPT was receiving 2.5 billion prompts per day.
August 2025: GPT-5 arrives
August was one of the most significant months of the year. OpenAI unveiled GPT-5, a next-generation model designed to be smarter and more useful. GPT-5 could handle tasks like coding apps, managing calendars, and creating research briefs, and it automatically decided whether to respond in a fast, thoughtful, or 'thinking' mode. The launch was not without issues; Altman acknowledged glitches and a 'router' problem that made the model seem dumber than expected at times.
OpenAI also released two open-weight models, gpt-oss-120b and gpt-oss-20b, marking a return to open-source AI after years of closed models. ChatGPT reached nearly 700 million weekly active users in early August, and the mobile app passed $2 billion in consumer spending since its 2023 launch. The company also offered ChatGPT Enterprise to federal agencies for $1 for a year, part of a broader push into government work.
September 2025: Shopping, Pulse, and parental controls
OpenAI made significant moves into e-commerce in September. The company launched Instant Checkout, allowing U.S. users to purchase products directly from Etsy and, soon, millions of Shopify merchants without leaving ChatGPT. This was a major step toward turning ChatGPT into a full shopping assistant. OpenAI also launched Pulse, a personalized morning briefing feature designed to make ChatGPT proactive and asynchronous.
On the safety front, OpenAI introduced parental controls for ChatGPT on web and mobile, following a tragic lawsuit over a teen's suicide. The company tightened rules for under-18 users, blocking flirtatious exchanges and adding stronger safeguards around conversations about suicide. OpenAI also rolled out GPT-5-Codex, an upgraded coding agent that could work for up to seven hours on complex tasks.
October 2025: The ChatGPT browser and 800 million users
OpenAI launched ChatGPT Atlas, an AI-powered browser, starting on Mac. Unlike traditional browsers, Atlas integrated ChatGPT into the search experience, challenging Google Chrome and AI search products. The browser was open to all users and was expected to arrive on Windows, iOS, and Android soon. Developers could also begin building interactive apps inside ChatGPT, with partners like Spotify, Expedia, Figma, and Canva.
By October, ChatGPT had surpassed 800 million weekly active users. OpenAI continued its shopping push by partnering with Walmart, and the ChatGPT Go plan, priced under $5, expanded to 16 more Asian countries.
November 2025: Group chats, voice integration, and legal battles
In November, OpenAI rolled out GPT-5.1, an upgraded version of GPT-5 with warmer and more conversational Instant mode, plus faster simple-task handling in Thinking mode. ChatGPT integrated voice mode directly into the main interface, eliminating the need to switch to a separate screen. Group chats also became available to all users, allowing friends, families, and coworkers to share a single conversation with ChatGPT.
The company also launched an AI shopping assistant ahead of the holiday season, helping users research products and make recommendations. OpenAI reached one million business clients, calling itself the fastest-growing business platform in history. However, legal troubles mounted. A Munich court ruled that ChatGPT violated German copyright law by reproducing lyrics from protected songs. Seven families filed lawsuits alleging that GPT-4o was released without adequate safeguards and contributed to suicides. OpenAI also faced a trademark lawsuit over the word 'cameo' in its Sora features.
December 2025: GPT-5.2, Disney, and a code red
December was a dramatic finish to the year. OpenAI released GPT-5.2, the latest version of its flagship model, with three variants: Instant, Thinking, and Pro. The company also rolled out GPT Image 1.5, a faster and more precise image generation model. Disney announced a three-year partnership with OpenAI, investing $1 billion and bringing hundreds of Disney, Marvel, Pixar, and Star Wars characters to Sora. The deal raised major questions about AI licensing and creative ownership, especially since Disney sued Google on the same day over copyright infringement.
Most notably, reports emerged that Altman had declared a 'code red' inside OpenAI, telling staff that the company would prioritize improving ChatGPT as competition from Google and other rivals intensified. As part of the shift, other initiatives, including advertising, were put on the back burner. OpenAI also added new tone controls for ChatGPT, updated guidelines to protect teen users, and highlighted an enterprise surge with ChatGPT message volume up eightfold since late 2024. The year ended with ChatGPT firmly at the center of OpenAI's strategy, but also facing unprecedented competition, scrutiny, and legal challenges.
Source: TechCrunch News