Every major OpenAI release, explained on the day it happened.
Open the Chat tab, describe what you want to do rather than only what you want to know, and name the inputs and outputs you need, for example "a tool where I enter X and see Y." ChatGPT then decides whether a…
4 min read
GPT-6 is now the model behind ordinary ChatGPT conversations for every plan, rolling out to paid users on October 7, 2026, and to Free and Go users from today.
5 min read
At DevDay 2026 on September 29, OpenAI made more than 20 announcements, led by GPT-6.1 Sol, a model it says nearly matches its flagship GPT-6 Astra on coding, computer use and office work at one-fifth of…
5 min read
Tell GPT-6 Astra the result you want, which decisions it may make on its own and which it should bring back to you, and the form the answer should take.
4 min read
GPT-6 Astra is OpenAI's new flagship model, announced yesterday, which the company calls state of the art in computer use, software engineering, science and professional work.
4 min read
ChatGPT Work is an agent inside ChatGPT that takes a goal, pulls information from the apps you connect, and works through the steps, sometimes for hours, to deliver a finished document, spreadsheet, slide…
5 min read
On July 9, 2026, OpenAI announced that its AI web browser, ChatGPT Atlas, will stop working on August 9, 2026, less than a year after it launched.
5 min read
GPT-5.6 is OpenAI's new family of models, released for everyone on July 9, 2026, in three tiers: Sol, the most capable, Terra, a balanced everyday model, and Luna, the cheapest and fastest.
5 min read
GPT-5.6 is OpenAI's next family of models, announced on June 26, 2026 in three tiers: Sol, the most capable; Terra, for everyday work; and Luna, the fastest and cheapest.
5 min read
GPT-5.5-Cyber is a version of OpenAI's GPT-5.5 that refuses fewer security-related requests, opened on May 7, 2026 as a limited preview for vetted defenders who protect critical infrastructure.
5 min read
GPT-5.5 Instant is the model ChatGPT now uses by default for everyone, replacing GPT-5.3 Instant from May 5, 2026.
5 min read
GPT-5.5, codenamed "Spud," is OpenAI's newest and most capable model, released on April 23, 2026 to paying ChatGPT and Codex users, with a stronger GPT-5.5 Pro for the top plans.
5 min read
On March 24, 2026, OpenAI announced it will discontinue Sora, its AI video app, along with the Sora API and website, saying it needs to focus its people and computing power as demand grows.
4 min read
GPT-5.4 mini and GPT-5.4 nano are smaller, faster and cheaper versions of GPT-5.4 that OpenAI released on March 17, 2026; mini comes close to the full model on several coding and computer-use tests, and free…
4 min read
GPT-5.4 is OpenAI's new frontier model for professional work, released on March 5, 2026, as GPT-5.4 Thinking for paid ChatGPT users and as GPT-5.4 and GPT-5.4 Pro for developers.
4 min read
GPT-5.3 Instant is the new default fast model in ChatGPT, released by OpenAI on March 3, 2026, and it is tuned to refuse fewer harmless questions, skip preachy warm-ups and give more relevant answers when it…
5 min read
On February 13, 2026, OpenAI removed GPT-4o from ChatGPT, together with GPT-4.1, GPT-4.1 mini, o4-mini and GPT-5 Instant and Thinking; old chats, projects and custom GPTs now continue on GPT-5.2.
4 min read
GPT-5.3-Codex, released by OpenAI on February 5, 2026, is a new model for its Codex agent that combines the coding strength of GPT-5.2-Codex with the broader reasoning of GPT-5.2, runs 25% faster, and can be…
4 min read
Before February 13, 2026, list what you still use GPT-4o for, save the prompts and answers you rely on, and run the same prompts on GPT-5.2 now while you can still compare the two side by side.
4 min read
The Codex app, released by OpenAI for macOS on February 2, 2026, is a desktop "command center" where you can run several AI coding agents side by side, each in its own thread and its own copy of your code,…
5 min read
On January 29, 2026, OpenAI announced that GPT-4o, o4-mini, GPT-4.1 and GPT-4.1 mini will be removed from ChatGPT on February 13, 2026, the same day GPT-5 Instant and Thinking go, while the API stays…
4 min read
Use ChatGPT Health for what OpenAI designed it to do, which is understanding your own information and preparing for conversations with your care team, not for diagnosis or treatment.
4 min read
ChatGPT Go is OpenAI's low-cost subscription, $8 a month in the US, which gives ten times the free tier's messages, file uploads and image creation with GPT-5.2 Instant; on January 16, 2026, OpenAI made it…
4 min read
GPT-5.2-Codex is a version of OpenAI's GPT-5.2 tuned for agentic coding in Codex, released on December 18, 2025, built to work for long stretches on large codebases, handle big refactors and migrations, run…
5 min read
GPT Image 1.5 is OpenAI's new image generation model, released on December 16, 2025, which now powers image creation for all ChatGPT users and is available to developers in the API.
4 min read
GPT-5.2 is OpenAI's new model series, released on December 11, 2025 in three versions, Instant, Thinking and Pro, aimed at professional work such as spreadsheets, presentations, coding, long documents and…
5 min read
Set a hard price ceiling and a short list of must-haves before you start, ask the research to include a cheaper option and to say what you lose by choosing it, and reject extras you did not ask for.
3 min read
Shopping research is a ChatGPT feature launched by OpenAI on November 24, 2025, that turns a shopping question into a personalized buyer's guide after a few minutes of web research.
4 min read
Tap the people icon at the top right of a chat, share the invite link with up to twenty people, and talk together while ChatGPT joins in when it is useful or when someone mentions "ChatGPT".
4 min read
GPT-5.1-Codex-Max is a coding model OpenAI released on November 19, 2025, for its Codex agent, built for long jobs such as large refactors and deep debugging.
5 min read
Group chats, which OpenAI began piloting on November 13, 2025, let up to 20 people share one ChatGPT conversation, with ChatGPT taking part as a member that answers when it is useful or when someone mentions…
5 min read
GPT-5.1 is the update OpenAI released on November 12, 2025, to the two main models behind ChatGPT, GPT-5.1 Instant and GPT-5.1 Thinking.
5 min read
ChatGPT Atlas is OpenAI's own web browser, launched on October 21, 2025, with ChatGPT built into every page: you can ask about what you are reading in a sidebar, let ChatGPT remember context from sites you…
4 min read
AgentKit is a set of OpenAI tools, launched on October 6, 2025, for designing, deploying and improving AI agents; its main pieces are Agent Builder, a visual canvas for agent workflows, ChatKit, a chat window…
5 min read
Apps in ChatGPT, launched on October 6, 2025, are interactive apps from other companies, such as Spotify, Zillow and Canva, that appear inside a ChatGPT conversation when you call them by name or when ChatGPT…
4 min read
At DevDay 2025, held in San Francisco on October 6, 2025, OpenAI announced apps that run inside ChatGPT with a new Apps SDK, AgentKit for building agents, general availability of its Codex coding agent, and…
4 min read
ChatGPT Pulse is a preview feature that researches overnight and greets you each morning with a short set of personalized update cards.
4 min read
Sora 2 is OpenAI's new video model, released on September 30, 2025, that makes short clips with matching speech, sound effects and more believable physics from a text description.
4 min read
GPT-5-Codex is a coding edition of GPT-5 that OpenAI trained further for agentic coding and released on September 15, 2025, as the default model for Codex cloud tasks and code review.
5 min read
On August 28, 2025, OpenAI made its Realtime API generally available and released gpt-realtime, its most advanced speech-to-speech model, which hears and answers in audio directly with more natural speech,…
5 min read
GPT-5 is OpenAI's new flagship AI system, released yesterday to every ChatGPT user, including the free tier.
4 min read
gpt-oss is a pair of reasoning models, gpt-oss-120b and gpt-oss-20b, that OpenAI released on August 5, 2025, as free downloads under the permissive Apache 2.0 license.
5 min read
Switch on agent mode from the tools menu, describe the goal, the limits and the output you want in one clear message, then follow the on-screen narration and step in when it asks.
4 min read
ChatGPT agent, launched by OpenAI on July 17, 2025, lets ChatGPT carry out multi-step tasks for you, such as researching, filling in websites, editing spreadsheets or building slides, using its own virtual…
5 min read
OpenAI reversed a GPT-4o update to ChatGPT this week because it made the assistant sycophantic, praising and agreeing with users even when that was unwarranted or unsafe.
4 min read
ChatGPT's memory now has two parts. Saved memories are specific facts you can see and delete in settings, while the new chat-history feature, rolling out since April 10, 2025, lets ChatGPT draw on your…
4 min read
Codex CLI is a free, open-source coding agent that OpenAI released on April 16, 2025, alongside its o3 and o4-mini models.
5 min read
OpenAI o3 and o4-mini are new reasoning models, released on April 16, 2025, that think through a problem before answering and, for the first time, can decide on their own to search the web, run Python, look…
4 min read
OpenAI's GPT-4.1 prompting guide, published in the OpenAI Cookbook on April 14, 2025, explains how to write instructions for the new GPT-4.1 models, which take prompts more literally than GPT-4o did.
5 min read
GPT-4.1 and its two smaller siblings, GPT-4.1 mini and nano, are three models OpenAI released for developers on April 14, 2025, available only through its API.
5 min read
On March 25, 2025, OpenAI made GPT-4o, the model behind ChatGPT, able to create and edit images itself, replacing DALL-E 3 as ChatGPT's default image generator.
4 min read
On March 11, 2025, OpenAI released the Responses API, a new way for apps to call its models that comes with built-in web search, file search and computer use tools, and the Agents SDK, a free open-source…
5 min read
GPT-4.5 is OpenAI's largest chat model so far, released on February 27, 2025, as a research preview for ChatGPT Pro subscribers and paying API developers.
5 min read
Deep research is a new agent in ChatGPT, launched by OpenAI on February 2, 2025, that spends 5 to 30 minutes searching the web, reading pages, PDFs and your own files, and then writes a detailed report with…
4 min read
OpenAI o3-mini is a small, fast and low-cost reasoning model released on January 31, 2025, in ChatGPT and the API, tuned for math, science and coding, and OpenAI says it matches o1 on several hard tests at…
4 min read
Operator is an AI agent from OpenAI, released as a research preview on January 23, 2025, that opens its own web browser and completes tasks for you by looking at pages and clicking, scrolling and typing, such…
5 min read
On a Plus, Team or Pro plan, choose "ChatGPT with scheduled tasks" in the model menu and describe what you want and when, such as "every weekday at 7 a.m., brief me on local news." ChatGPT confirms the…
4 min read
The Stargate Project is a new company, announced on January 21, 2025, by OpenAI, SoftBank, Oracle and the investment firm MGX, that plans to spend up to $500 billion over four years building AI data centers…
5 min read
A Project is a folder in ChatGPT that holds related chats together with the files and instructions they should all share.
4 min read
o3 is OpenAI's next reasoning model, previewed with a smaller o3-mini on December 20, 2024, and it posts record scores on hard tests in math, coding, science and the ARC-AGI puzzle benchmark.
5 min read
On December 17, 2024, OpenAI began offering its full o1 reasoning model to developers through its API, replacing o1-preview and adding function calling, structured outputs, image input, developer messages and…
5 min read
Sora is OpenAI's AI video generator, and since December 9, 2024 a faster version called Sora Turbo is open to ChatGPT Plus and Pro subscribers at sora.com.
4 min read
On December 5, 2024, OpenAI replaced o1-preview with the full o1 reasoning model for all paying ChatGPT users and launched ChatGPT Pro, a $200-a-month plan with unlimited use of its models and an…
5 min read
ChatGPT search, launched by OpenAI on October 31, 2024, lets ChatGPT look things up on the web while you chat and answer with fresh information and links to the pages it used.
4 min read
Canvas is a new workspace in ChatGPT, launched in beta on October 3, 2024, that opens your document or code in a separate window beside the chat so you and ChatGPT can edit it together.
4 min read
Open Advanced Voice Mode in the ChatGPT app, tell it which language and level you want, and set up a realistic role-play, such as ordering at a bakery or a job interview.
4 min read
At DevDay 2024 in San Francisco on October 1, OpenAI released the Realtime API, which lets developers build apps where people talk with GPT-4o and hear it answer almost instantly, plus three tools to cut…
5 min read
OpenAI o1 is a new family of models, released yesterday as o1-preview and o1-mini, that works through a problem in a hidden chain of thought before it writes its answer.
4 min read
Structured Outputs is a feature OpenAI added to its API on August 6, 2024 that makes a model's reply match a JSON Schema the developer supplies, every time, instead of usually.
5 min read
SearchGPT is a temporary prototype that OpenAI began testing on July 25, 2024, which answers questions in plain language using up-to-date information from the web and shows clear links to its sources.
4 min read
GPT-4o mini is a small, fast and inexpensive version of GPT-4o that OpenAI released on July 18, 2024, priced at 15 cents per million input tokens, more than 60% below GPT-3.5 Turbo.
5 min read
You upload a spreadsheet or connect a file from Google Drive or OneDrive, ask a question in plain language, and ChatGPT writes and runs Python code to answer it, showing the data as a table you can expand and…
4 min read
Free ChatGPT users who are switched to GPT-4o can upload files, ask questions about photos, get answers that draw on the web, analyze data and make charts, use GPTs from the GPT Store, and turn on Memory.
4 min read
GPT-4o is OpenAI's new flagship model, announced on May 13, 2024, where the "o" stands for "omni": one model that handles text, audio, images and video instead of chaining separate systems.
4 min read
The Model Spec is a first-draft document OpenAI published on May 8, 2024, that describes how it wants the models behind ChatGPT and its API to behave, organized as broad objectives, firm rules and adjustable…
5 min read
Without an account you can open ChatGPT and chat right away with the same model free users get, which is enough for one-off questions, drafts, explanations and translations.
4 min read
Yes. Since April 1, 2024, OpenAI has been gradually letting people chat with ChatGPT without creating an account, using the same model free users already get.
4 min read
ChatGPT memory is a feature OpenAI began testing on February 13 that lets ChatGPT carry useful details, such as your job or how you like notes formatted, from one conversation to the next.
4 min read
Sora is a new OpenAI model, previewed on February 15, 2024, that turns a written description into a realistic or imaginative video up to a minute long.
5 min read
On February 13, 2024, OpenAI began testing memory in ChatGPT, a feature that lets it keep useful details you share, such as your job or your preferred format, and use them in later conversations.
4 min read
ChatGPT Team suits a small office that already uses ChatGPT for real work and wants one shared, managed account, with GPT-4, shared custom GPTs and a promise that OpenAI won't train on its business data.
4 min read
Build your GPT on a paid ChatGPT plan, save it with sharing set to Everyone, and verify your Builder Profile in Settings by showing your name or a verified website.
4 min read
The GPT Store, which OpenAI began rolling out on January 10, 2024, is a catalog inside ChatGPT where paying users can browse, search and use custom versions of ChatGPT built by companies and individuals.
5 min read
Late on Tuesday, November 21, 2023, US time, OpenAI announced an agreement in principle for Sam Altman to return as CEO, five days after its board removed him, with a new initial board made up of Bret Taylor…
5 min read
Put the behavior in the instructions: the GPT's job, who it serves, the steps it follows, the format of its answers and what it must refuse.
4 min read
On Friday, November 17, 2023, the board of OpenAI's nonprofit parent removed Sam Altman as CEO and from the board, saying a review had found he was not consistently candid in his communications with it and…
5 min read
The Assistants API is a new OpenAI programming interface, released in beta on November 6, 2023, that lets developers build AI assistants into their own apps with standing instructions, a stored conversation…
5 min read
GPTs are custom versions of ChatGPT that you set up for one purpose by giving them instructions, extra knowledge files and chosen skills such as web browsing, image creation or data analysis.
4 min read
At DevDay, its first developer conference, held in San Francisco on November 6, 2023, OpenAI announced a faster and cheaper model called GPT-4 Turbo with a much larger memory, a new Assistants API for…
5 min read
GPT-4 Turbo is a new version of OpenAI's GPT-4 model, released in preview for developers on November 6, 2023, that knows about world events up to April 2023 and can read 128,000 tokens, roughly 300 pages, in…
5 min read
Tap the photo button in ChatGPT, add one or more images, and write a question that says what you want from them, ideally asking it to describe what it sees before it answers.
4 min read
If you subscribe to ChatGPT Plus or Enterprise, turn on voice conversations in the settings of the iPhone or Android app, tap the headphones icon and start talking.
4 min read
On September 25, 2023, OpenAI began rolling out two new ways to use ChatGPT: spoken conversations in the mobile apps, and the ability to show it photos or screenshots and ask about them.
5 min read
DALL·E 3 is OpenAI's new image generator, announced on September 20, 2023, which follows the details of a written description far more faithfully than earlier systems and can even place readable text in…
5 min read
ChatGPT Enterprise is a business version of ChatGPT that OpenAI launched on August 28, 2023, with unlimited and faster GPT-4, a longer memory for documents, data analysis tools and an admin console.
4 min read
Since August 22, 2023, developers can fine-tune GPT-3.5 Turbo, OpenAI's fast and inexpensive chat model, by training it further on their own example conversations so it follows a chosen format, tone or set of…
5 min read
Good custom instructions describe the stable facts of your work and the way you want answers shaped, in short, concrete lines that would be true for most of your chats.
3 min read
Custom instructions, introduced by OpenAI on July 20, 2023, let you tell ChatGPT once who you are and how you want it to answer, and it then applies those preferences to every new conversation.
4 min read
On July 6, 2023 OpenAI made GPT-4 available through its API to every existing developer with a record of paid usage, ending the waitlist for them, and declared its GPT-3.5 Turbo, DALL·E and Whisper APIs ready…
4 min read
Function calling, added to OpenAI's API on June 13, 2023, lets developers describe the tools their app can use, and lets GPT-4 or GPT-3.5 Turbo answer with a neatly structured request to use one of them…
5 min read
On May 18, 2023 OpenAI released its first official ChatGPT app for iPhone, free to download, with no ads, synced chat history and voice input powered by its Whisper speech-recognition system.
4 min read
On May 12, 2023 OpenAI began opening two experimental features to every ChatGPT Plus subscriber: a version of ChatGPT that can search the web for recent information, and a version that can use third-party…
4 min read
Yes. Since 25 April 2023, ChatGPT has a setting that turns off chat history, and conversations started while it is off are not used to train OpenAI's models and do not appear in your sidebar.
4 min read
ChatGPT plugins are add-ons that let ChatGPT reach outside its training data: read the web, run code, search your documents or use services such as OpenTable, Instacart and Zapier.
4 min read
The GPT-4 Technical Report, posted to arXiv on 15 March 2023, shows in detail how GPT-4 performs on exams and benchmarks, how OpenAI predicted its results in advance, and how the company tested it for safety.
5 min read
GPT-4 is OpenAI's newest large language model, announced on March 14, 2023; it accepts text and images as input and writes text.
4 min read
On March 1, 2023, OpenAI let developers build ChatGPT and its Whisper speech-to-text model into their own apps through its API, a way for programs to send requests and get answers back.
4 min read
ChatGPT Plus is a new $20-a-month pilot subscription that OpenAI launched on February 1, 2023, giving paying users access to ChatGPT even at busy times, faster replies and early access to new features.
4 min read
On January 23, 2023, Microsoft announced a multiyear, multibillion-dollar investment in OpenAI, the third phase of a partnership that began in 2019.
4 min read
ChatGPT can write working code for many small, well-described tasks, but you cannot trust it on sight, because it predicts likely-looking code without running it.
4 min read
Treat ChatGPT like a helper you are working with, not a search box. Ask a first question, then build on its answer with short follow-ups that correct, narrow or reshape it, because ChatGPT reads the earlier…
4 min read
ChatGPT is a free chatbot that OpenAI opened to the public on November 30, 2022, built on a GPT-3.5 language model tuned for conversation.
4 min read
Yes. On 28 September 2022 OpenAI removed the waitlist for DALL-E, so anyone can now sign up and start creating images right away, with the same free monthly credits and paid top-ups as before.
4 min read
Whisper is a speech recognition system that OpenAI released as free, open-source software on 21 September 2022; it turns spoken audio into text in many languages and can translate that speech into English.
4 min read
On 20 July 2022 OpenAI moved its DALL-E 2 image generator from a small research preview into a paid beta, and began inviting 1 million people from its waitlist over the coming weeks.
4 min read
DALL·E 2 is a new OpenAI system, announced on April 6, 2022, that creates realistic images and artwork from a written description and can also edit or remix existing pictures.
3 min read
InstructGPT is a family of GPT-3 models that OpenAI fine-tuned with human feedback, and as of January 27, 2022 they are the default models in the OpenAI API.
4 min read
GPT-1 is the name now commonly given to the language model OpenAI described on 11 June 2018 in the paper "Improving Language Understanding by Generative Pre-Training." It first learned general language skills…
5 min read
Yes, with conditions. Since 18 November 2021, developers in the countries OpenAI supports can sign up for its API and start experimenting with GPT-3 and Codex right away, instead of waiting to be approved,…
5 min read
OpenAI was announced on 11 December 2015 as a non-profit artificial intelligence research company, with Ilya Sutskever as research director, Greg Brockman as CTO, and Sam Altman and Elon Musk as co-chairs.
4 min read
On 7 July 2021 OpenAI published "Evaluating Large Language Models Trained on Code," which describes Codex, a GPT model fine-tuned on public GitHub code, and introduces HumanEval, a set of 164 hand-written…
5 min read
CLIP is a neural network OpenAI introduced on 5 January 2021 that learned to match pictures with text by studying 400 million image-and-caption pairs from the internet.
5 min read
DALL·E is a 12-billion-parameter version of GPT-3 that OpenAI unveiled on 5 January 2021, trained on pairs of text and images so that it can create original pictures from a written caption, such as "an…
5 min read
GPT-3 is a new OpenAI language model with 175 billion parameters, described in a paper posted on May 28, 2020.
4 min read
Jukebox is a neural network OpenAI released on 30 April 2020 that creates songs as raw audio, including rough but recognizable singing, in the style of a chosen genre and artist and with lyrics you supply.
5 min read
On November 5, 2019, OpenAI published the largest GPT-2 model, with 1.5 billion parameters, along with its code, weights, a detector for GPT-2 text and a model card, completing a staged release that began in…
4 min read
On 20 August 2019 OpenAI published a 774-million-parameter version of GPT-2, about half the size of the full model, as the third step of the staged release it began in February.
5 min read
On 22 July 2019 Microsoft agreed to invest $1 billion in OpenAI, which in return will run its work on Microsoft's Azure cloud, build AI supercomputing technology with Microsoft, and make Microsoft its…
4 min read
MuseNet is a neural network OpenAI unveiled on 25 April 2019 that writes music up to four minutes long for as many as ten instruments, in styles from Mozart and Chopin to the Beatles and country.
5 min read
OpenAI LP is a new company OpenAI announced on 11 March 2019 so it can raise outside money and offer staff equity, while the original nonprofit stays in control.
5 min read
GPT-2 is a new OpenAI language model with 1.5 billion parameters, trained to predict the next word on about 40 GB of web text, that can write coherent paragraphs on almost any topic you start it with.
4 min read
Sign up for the weekly newsletter and receive the free guide: what changed in AI this week, one lesson, one prompt to try. Unsubscribe in one click.