\n<\/p>\n
ChatGPT, OpenAI\u2019s text-generating AI chatbot, has taken the world by storm since its launch in November 2022.<\/a> What started as a tool to supercharge productivity through writing essays and code with short text prompts<\/a> has evolved into a behemoth with 300 million weekly active users.<\/a><\/p>\n 2024 was a big year for OpenAI, from its partnership with Apple<\/a> for its generative AI offering, Apple Intelligence,<\/a> the release of GPT-4o with voice capabilities,<\/a> and the highly-anticipated launch of its text-to-video model Sora.<\/a><\/p>\n OpenAI also faced its share of internal drama, including the notable exits of high-level execs like co-founder and longtime chief scientist Ilya Sutskever<\/a> and CTO Mira Murati.<\/a> OpenAI has also been hit with lawsuits from Alden Global Capital-owned newspapers<\/a> alleging copyright infringement, as well as an injunction from Elon Musk<\/a> to halt OpenAI\u2019s transition to a for-profit.<\/p>\n In 2025, OpenAI is battling the perception that it\u2019s ceding ground in the AI race to Chinese rivals like DeepSeek<\/a>. The company has been trying to shore up its relationship with Washington<\/a> as it simultaneously pursues an ambitious data center project,<\/a> and as it reportedly lays the groundwork<\/a> for one of the largest funding rounds in history.<\/p>\n Below, you\u2019ll find a timeline of ChatGPT product updates and releases, starting with the latest, which we\u2019ve been updating throughout the year. If you have any other questions, check out our ChatGPT FAQ here.<\/a><\/p>\n Noyb, a privacy rights advocacy group, is supporting an individual in Norway who was shocked to discover that ChatGPT<\/a> was providing false information about him, stating that he had been found guilty of killing two of his children and trying to harm the third. \u201cThe GDPR is clear. Personal data has to be accurate,\u201d said Joakim S\u00f6derberg, data protection lawyer at\u00a0Noyb, in a statement. \u201cIf it\u2019s not, users have the right to have it changed to reflect the truth. Showing ChatGPT users a tiny disclaimer that the chatbot can make mistakes clearly isn\u2019t enough. You can\u2019t just spread false information and in the end add a small disclaimer saying that everything you said may just not be true.\u201d<\/p>\n OpenAI has added new transcription and voice-generating AI models to its APIs<\/a>: a text-to-speech model, \u201cgpt-4o-mini-tts,\u201d that delivers more nuanced and realistic sounding speech, as well as two speech-to-text models called \u201cgpt-4o-transcribe\u201d and \u201cgpt-4o-mini-transcribe\u201d. The company claims they are improved versions of what was already there and that they hallucinate less.<\/p>\n OpenAI<\/a> has introduced o1-pro<\/a> in its developer API. OpenAI says its o1-pro uses more computing than its o1 \u201creasoning\u201d AI model<\/a> to deliver \u201cconsistently better responses.\u201d It\u2019s only accessible to select developers who have spent at least $5 on OpenAI API services. OpenAI charges $150 for every million tokens (about 750,000 words) input into the model and $600 for every million tokens the model produces. It costs twice as much as OpenAI\u2019s GPT-4.5<\/a> for input and 10 times the price of regular o1.<\/p>\n Noam Brown, who heads AI reasoning research at OpenAI, thinks that certain types of AI models for \u201creasoning\u201d could have been developed 20 years ago<\/a> if researchers had understood the correct approach and algorithms.<\/p>\n OpenAI CEO Sam Altman said, in a\u00a0post on X<\/a>, that the company has trained a \u201cnew model\u201d that\u2019s \u201creally good\u201d at creative writing<\/a>. He posted a lengthy sample from the model given the prompt \u201cPlease write a metafictional literary short story about AI and grief.\u201d OpenAI has not extensively explored the use of AI for writing fiction. The company has mostly concentrated on challenges in rigid, predictable areas such as math and programming.\u00a0<\/em>And it turns out that it might not be that great<\/a> at creative writing at all. <\/p>\n we trained a new model that is good at creative writing (not sure yet how\/when it will get released). this is the first time i have been really struck by something written by AI; it got the vibe of metafiction so right.<\/p>\n PROMPT:<\/p>\n Please write a metafictional literary short story\u2026<\/p>\n \u2014 Sam Altman (@sama) March 11, 2025<\/a><\/p><\/blockquote>\n OpenAI rolled out new tools <\/a>designed to help developers and businesses build AI agents<\/a> \u2014 automated systems that can independently accomplish tasks \u2014 using the company\u2019s own AI models and frameworks. The tools are part of OpenAI\u2019s new Responses API, which enables enterprises to develop customized AI agents that can perform web searches, scan through company files, and navigate websites, similar to OpenAI\u2019s Operator product<\/a>. The Responses API effectively replaces OpenAI\u2019s Assistants API<\/a>, which the company plans to discontinue in the first half of 2026.<\/p>\n OpenAI intends to release several \u201cagent\u201d products<\/a> tailored for different applications, including sorting and ranking sales leads and software engineering, according to a report from The Information<\/a>. One, a \u201chigh-income knowledge worker\u201d agent, will reportedly be priced at $2,000 a month. Another, a software developer agent, is said to cost $10,000 a month. The most expensive rumored agents, which are said to be aimed at supporting \u201cPhD-level research,\u201d are expected to cost $20,000 per month. The jaw-dropping figure is indicative of how much cash OpenAI needs right now: The company lost roughly $5 billion last year after paying for costs related to running its services and other expenses. It\u2019s unclear when these agentic tools might launch or which customers will be eligible to buy them.<\/p>\n The latest version of the macOS ChatGPT app allows users to edit code directly in supported developer tools<\/a>, including Xcode, VS Code, and JetBrains. ChatGPT Plus, Pro, and Team subscribers can use the feature now, and the company plans to roll it out to more users like Enterprise, Edu, and free users.<\/p>\n According to a new report<\/a> from VC firm Andreessen Horowitz (a16z), OpenAI\u2019s AI chatbot, ChatGPT, experienced solid growth in the second half of 2024.<\/a> It took ChatGPT nine months to increase its weekly active users from 100 million in November 2023<\/a> to 200 million in August 2024, but it only took less than six months to double that number once more, according to the report. ChatGPT\u2019s weekly active users increased to 300 million by December 2024<\/a> and 400 million by February 2025<\/a>. ChatGPT has experienced significant growth recently due to the launch of new models and features, such as GPT-4o, with multimodal capabilities. ChatGPT usage spiked from April to May 2024, shortly after that model\u2019s launch.<\/p>\n OpenAI has effectively canceled the release of o3 in favor of what CEO Sam Altman is calling a \u201csimplified\u201d product offering. In a post on X<\/a>, Altman said that, in the coming months, OpenAI will release a model called GPT-5 that \u201cintegrates a lot of [OpenAI\u2019s] technology,\u201d including o3, in ChatGPT and its API. As a result of that roadmap decision, OpenAI no longer plans to release o3 as a standalone model.\u00a0<\/p>\n A commonly cited stat<\/a> is that ChatGPT requires around 3 watt-hours of power to answer a single question. Using OpenAI\u2019s latest default model for ChatGPT, GPT-4o, as a reference, nonprofit AI research institute Epoch AI<\/a> found the average ChatGPT query consumes around 0.3 watt-hours.<\/a> However, the analysis doesn\u2019t consider the additional energy costs incurred by ChatGPT with features like image generation or input processing.<\/p>\n In response to pressure from rivals like DeepSeek, OpenAI is changing the way its o3-mini model communicates its step-by-step \u201cthought\u201d process.<\/a> ChatGPT users will see an updated \u201cchain of thought\u201d that shows more of the model\u2019s \u201creasoning\u201d steps and how it arrived at answers to questions.<\/p>\n OpenAI is now allowing anyone to use ChatGPT web search without having to log in.<\/a> While OpenAI had previously allowed<\/a> users to ask ChatGPT questions without signing in, responses were restricted to the chatbot\u2019s last training update. This only applies through ChatGPT.com, however. To use ChatGPT in any form through the native mobile app, you will still need to be logged in.<\/p>\n OpenAI announced a new AI \u201cagent\u201d called deep research<\/a> that\u2019s designed to help people conduct in-depth, complex research using ChatGPT. OpenAI says the \u201cagent\u201d is intended for instances where you don\u2019t just want a quick answer or summary, but instead need to assiduously consider information from multiple websites and other sources.<\/p>\n OpenAI used the subreddit r\/ChangeMyView to measure the persuasive abilities<\/a> of its AI reasoning models. OpenAI says it collects user posts from the subreddit<\/a> and asks its AI models to write replies, in a closed environment, that would change the Reddit user\u2019s mind on a subject. The company then shows the responses to testers, who assess how persuasive the argument is, and finally OpenAI compares the AI models\u2019 responses to human replies for that same post.\u00a0<\/p>\n OpenAI launched a new AI \u201creasoning\u201d model, o3-mini,<\/a> the newest in the company\u2019s o family of models. OpenAI first previewed the model in December<\/a> alongside a more capable system called o3. OpenAI is pitching its new model as both \u201cpowerful\u201d and \u201caffordable.\u201d<\/p>\n A new report from app analytics firm Appfigures found that over half of ChatGPT\u2019s mobile users<\/a> are under age 25, with users between ages 50 and 64 making up the second largest age demographic. The gender gap among ChatGPT users is even more significant. Appfigures estimates that across age groups, men make up 84.5% of all users.<\/p>\n OpenAI launched ChatGPT Gov designed to provide U.S. government agencies<\/a> an additional way to access the tech. ChatGPT Gov includes many of the capabilities found in OpenAI\u2019s corporate-focused tier, ChatGPT Enterprise. OpenAI says that ChatGPT Gov enables agencies to more easily manage their own security, privacy, and compliance, and could expedite internal authorization of OpenAI\u2019s tools for the handling of non-public sensitive data.<\/p>\n Younger Gen Zers are embracing ChatGPT, for schoolwork, according to a new survey by the Pew Research Center.<\/a> In a follow-up to its 2023 poll on ChatGPT usage among young people, Pew asked ~1,400 U.S.-based teens ages 13 to 17 whether they\u2019ve used ChatGPT for homework or other school-related assignments. Twenty-six percent said that they had, double the number two years ago. Just over half of teens responding to the poll said they think it\u2019s acceptable to use ChatGPT for researching new subjects. But considering the ways ChatGPT can fall short, the results are possibly cause for alarm.<\/p>\n OpenAI says that it might store chats and associated screenshots from customers who use Operator, the company\u2019s AI \u201cagent\u201d tool,<\/a> for up to 90 days \u2014 even after a user manually deletes them. While OpenAI has a similar deleted data retention policy for ChatGPT, the retention period for ChatGPT is only 30 days, which is 60 days shorter than Operator\u2019s.<\/a><\/p>\n OpenAI is launching a research preview of Operator, a general-purpose AI agent that can take control of a web browser and independently perform certain actions.<\/a> Operator promises to automate tasks such as booking travel accommodations, making restaurant reservations, and shopping online.<\/p>\n Operator, OpenAI\u2019s agent tool,<\/a> could be released sooner rather than later. Changes to ChatGPT\u2019s code base suggest that Operator will be available as an early research preview<\/a> to users on the $200 Pro subscription plan. The changes aren\u2019t yet publicly visible, but a user on X who goes by Choi<\/a> spotted these updates in ChatGPT\u2019s client-side code. TechCrunch separately identified the same references to Operator on OpenAI\u2019s website.<\/p>\n OpenAI has begun testing a feature that lets new ChatGPT users sign up with only a phone number<\/a> \u2014 no email required. The feature is currently in beta in the U.S. and India. However, users who create an account using their number can\u2019t upgrade to one of OpenAI\u2019s paid plans without verifying their account via an email. Multi-factor authentication also isn\u2019t supported without a valid email.<\/p>\n ChatGPT\u2019s new beta feature, called tasks, allows users to set simple reminders<\/a>. For example, you can ask ChatGPT to remind you when your passport expires in six months, and the AI assistant will follow up with a push notification on whatever platform you have tasks enabled. The feature will start rolling out to ChatGPT Plus, Team, and Pro users around the globe this week.<\/p>\n OpenAI is introducing a new way for users to customize their interactions with ChatGPT.<\/a> Some users found they can specify a preferred name or nickname and \u201ctraits\u201d they\u2019d like the chatbot to have. OpenAI suggests traits like \u201cChatty,\u201d \u201cEncouraging,\u201d and \u201cGen Z.\u201d However, some users reported<\/a> that the new options have disappeared, so it\u2019s possible they went live prematurely.<\/p>\n ChatGPT Search can be fooled into generating completely misleading summaries, The Guardian has found.<\/a> They found ChatGPT could be prompted to ignore negative reviews and generate \u201centirely positive\u201d summaries<\/a> by inserting hidden text into websites it created and that ChatGPT Search could also be made to spit out malicious code using this method.<\/p>\n Microsoft and OpenAI have a very specific, internal definition of AGI<\/a> based on the startup\u2019s profits, according to a new report from The Information.<\/a> The two companies reportedly signed an agreement stating OpenAI has only achieved AGI when it develops AI systems that can generate at least $100 billion in profit, which is far from the rigorous technical and philosophical definition of AGI many would expect.<\/p>\n OpenAI released new research<\/a> outlining the company\u2019s approach to ensure AI reasoning models stay aligned with the values of their human developers. The startup used \u201cdeliberative alignment\u201d to make o1 and o3 \u201cthink\u201d about OpenAI\u2019s safety policy.<\/a> According to OpenAI\u2019s research, the method decreased the rate at which o1 answered \u201cunsafe\u201d questions while improving its ability to answer benign ones.<\/p>\n OpenAI CEO Sam Altman announced the successors to its o1 reasoning model family:<\/a> o3 and o3-mini. The models are not widely available yet, but safety researchers can sign up for a preview. The reveal marks the end of the \u201c12 Days of OpenAI\u201d event, which saw announcements for real-time vision capabilities, ChatGPT Search, and even a Santa voice for ChatGPT.\u00a0<\/p>\n In an effort to make ChatGPT accessible to as many people as possible, OpenAI announced a 1-800 number to call the chatbot<\/a> \u2014 even from a landline or a flip phone. Users can call 1-800-CHATGPT, and ChatGPT will respond to your queries in an experience that is more or less identical to Advanced Voice Mode \u2014 minus the multimodality.<\/p>\n OpenAI is offering 15 minutes of free calling for U.S. users. The company notes that standard carrier fees may apply.<\/p>\n OpenAI is bringing ChatGPT Search to free, logged in users.<\/a> Search gives ChatGPT the ability to access real-time information on the web to better answer your queries, but was only available for paid users when it launched in October.<\/a> Not only is Search available now for free users, but it\u2019s also been integrated into Advanced Voice Mode.<\/a><\/p>\n OpenAI is blaming one of the longest outages in its history on a \u201cnew telemetry service\u201d gone awry. OpenAI wrote in a postmortem that the outage wasn\u2019t caused by a security incident or recent product launch, but by a telemetry service it deployed<\/a> to collect Kubernetes metrics.<\/p>\n OpenAI announced that ChatGPT users could access a new \u201cSanta Mode\u201d voice<\/a> during December. The feature allows users to speak with ChatGPT\u2019s Advanced Voice Mode, but with a Christmas twist. The voice sounds, well, \u201cmerry and bright,\u201d as OpenAI described it. Think boomy, jolly \u2014 more or less like every Santa you\u2019ve ever heard.<\/p>\n OpenAI released the real-time video capabilities for ChatGPT<\/a> that it demoed nearly seven months ago. ChatGPT Plus, Team, and Pro subscribers can use the app to point their phones at objects and have ChatGPT respond in near-real-time. The feature can also understand what\u2019s on a device\u2019s screen through screen sharing.<\/p>\n There\u2019s more to come from OpenAI through December 23. Tune in to our live blog<\/a> to stay updated.<\/p>\n ChatGPT and Sora both experienced a major outage Wednesday. Though users suspected the outage was due to the rollout of ChatGPT in Apple Intelligence, OpenAI developer community lead Edwin Arbus denied it in a post on X<\/a>, saying the \u201coutage was unrelated to 12 Days of OpenAI or Apple Intelligence. We made a config change that caused many servers to become unavailable.\u201d<\/p>\n Canvas, OpenAI\u2019s collaboration-focused interface<\/a> for writing and code projects, is now rolling out to all users after being in beta for ChatGPT Plus members since October 2024. The company also announced the ability to integrate Python code within Canvas as well as bringing Canvas to custom GPTs.<\/p>\n OpenAI CEO Sam Altman posted on X that due to higher than expected demand, they are pausing new sign-ups for its video generator Sora and that video generations will be slower for the time being. The company released Sora<\/a> as part of its \u201c12 Days of OpenAI\u201d event following nearly a year of teasing the product.<\/a><\/p>\n OpenAI has finally released its text to video model, Sora.<\/a> The model can generate videos up to 20 seconds long in 1080p based on text prompts or uploaded images, and can be \u201cremixed\u201d through additional user prompts. Sora is available starting today to ChatGPT Pro<\/a> and Plus subscribers (except in the EU).\u00a0<\/p>\n In Monday\u2019s \u201c12 Days of OpenAI\u201d livestream,<\/a> CEO Sam Altman said that ChatGPT Plus members will get 50 video generations a month, while ChatGPT Pro users will get \u201cunlimited\u201d generations in their \u201cslow queue mode\u201d and 500 \u201cnormal\u201d generations per month.<\/p>\n There are still more reveals to come from OpenAI through December 23. Tune in to our live blog<\/a> to stay updated.<\/p>\n On day one of its 12 Days of OpenAI event,<\/a> the company announced a new \u2014 and expensive \u2014 subscription plan. ChatGPT Pro is a $200-per-month tier<\/a> that provides unlimited access to all of OpenAI\u2019s models, including the full version of its o1 \u201creasoning\u201d model.\u00a0<\/p>\n The full version of o1, which was released as a preview in September,<\/a> can now reason about image uploads and has been trained to be \u201cmore concise in its thinking\u201d to improve response times.\u00a0<\/p>\n Over the next few weeks, we\u2019ll be updating all the news from OpenAI as it happens on our live blog.<\/a> Follow along with us!<\/p>\nTimeline of the most recent ChatGPT updates<\/h2>\n
March 2025<\/h3>\n
OpenAI faces privacy complaint in Europe for chatbot\u2019s defamatory hallucinations<\/h4>\n
OpenAI upgrades its transcription and voice-generating AI models<\/h4>\n
OpenAI has launched o1-pro, a more powerful version of its o1<\/h4>\n
OpenAI research lead Noam Brown thinks AI \u201creasoning\u201d models could\u2019ve arrived decades ago<\/h4>\n
OpenAI says it has trained an AI that\u2019s \u201creally good\u201d at creative writing<\/h4>\n
\n
OpenAI launches new tools to help businesses build AI agents<\/h4>\n
OpenAI reportedly plans to charge up to $20,000 a month for specialized AI \u2018agents\u2019<\/h4>\n
ChatGPT can directly edit your code<\/strong><\/h4>\n
ChatGPT\u2019s weekly active users doubled in less than 6 months, thanks to new releases<\/h4>\n
February 2025 <\/h3>\n
OpenAI cancels its o3 AI model in favor of a \u2018unified\u2019 next-gen release<\/h4>\n
ChatGPT may not be as power-hungry as once assumed<\/h4>\n
OpenAI now reveals more of its o3-mini model\u2019s thought process<\/h4>\n
You can now use ChatGPT web search without logging in<\/h4>\n
OpenAI unveils a new ChatGPT agent for \u2018deep research\u2019<\/h4>\n
February 2025<\/strong><\/h3>\n
OpenAI used a subreddit to test AI persuasion<\/h4>\n
OpenAI launches o3-mini, its latest \u2018reasoning\u2019 model<\/h4>\n
ChatGPT\u2019s mobile users are 85% male, report says<\/h4>\n
OpenAI launches ChatGPT plan for US government agencies<\/h4>\n
More teens report using ChatGPT for schoolwork, despite the tech\u2019s faults<\/h4>\n
OpenAI says it may store deleted Operator data for up to 90 days<\/h4>\n
OpenAI launches Operator, an AI agent that performs tasks autonomously<\/h4>\n
OpenAI may preview its agent tool for users on the $200-per-month Pro plan<\/h4>\n
OpenAI tests phone number-only ChatGPT signups<\/h4>\n
ChatGPT now lets you schedule reminders and recurring tasks<\/h4>\n
New ChatGPT feature lets users assign it traits like \u2018chatty\u2019 and \u2018Gen Z\u2019<\/h4>\n
December 2024<\/h3>\n
ChatGPT Search can be tricked into misleading users, new research reveals<\/h4>\n
Microsoft and OpenAI reportedly have a finance-centric definition of AGI<\/h4>\n
OpenAI trained o1 and o3 to \u2018think\u2019 about its safety policy<\/h4>\n
OpenAI announces new o3 reasoning models<\/h4>\n
OpenAI brings ChatGPT to your landline<\/h4>\n
OpenAI brings its ChatGPT Search to more users<\/h4>\n
OpenAI blames massive ChatGPT outage on a \u2018new telemetry service\u2019<\/h4>\n
You can make ChatGPT sound like Santa for a limited time<\/h4>\n
OpenAI adds vision to Advanced Voice Mode<\/h4>\n
ChatGPT and Sora hit with a major outage<\/h4>\n
Canvas rolls out to everyone<\/h4>\n
OpenAI pauses Sora sign-ups due to high demand<\/h4>\n
OpenAI releases Sora for ChatGPT Plus and Pro subscribers<\/h4>\n
OpenAI launches $200 monthly ChatGPT Pro subscription \u2014 and full version of o1<\/h4>\n
OpenAI announces 12 days of reveals for the holidays<\/h4>\n