What is OpenAI? Complete History: GPT-6.1 Sol, Dots, ChatGPT Release Date & More (2026)

OpenAI is an AI research and deployment company founded in December 2015 in San Francisco, and it made ChatGPT, which came out on November 30, 2022. It began as a nonprofit lab, now runs as a public benefit corporation controlled by the OpenAI Foundation, and makes the GPT model series, Codex and dots. In just over a decade it went from $1 billion in pledges to an $852 billion valuation and a billion weekly ChatGPT users - the most dramatic rise in tech history.

But behind the record-breaking growth lies a story of boardroom coups, mission drift, $8.5 billion in annual losses, and a CEO who was fired and rehired within five days. In this article, we trace the complete history of OpenAI - from Turing's first spark to GPT-6, the September 29, 2026 DevDay launch of dots and GPT-6.1 Sol, and the race to superintelligence. 🔮

TL;DR: ChatGPT came out on November 30, 2022, and OpenAI grew from a 2015 nonprofit into an $852B public benefit corporation. At DevDay on September 29, 2026, it launched dots, always-on agents powered by GPT-6 Astra, and GPT-6.1 Sol, which scores one point below Astra at one-fifth of the price. Build with frontier models →

📅 Key Dates: When Did OpenAI and ChatGPT Launch?

ChatGPT came out on November 30, 2022, when OpenAI released it as a free research preview built on GPT-3.5, and OpenAI itself was founded in December 2015 as a nonprofit research lab. The table below answers the launch-date questions people search for most, with the newest GPT-6 milestones at the bottom.

One date needs a caveat. OpenAI launched GPT-6 Astra on September 3, 2026 as a limited rollout and said wider access would follow "over the coming days," and its launch page gives no exact general-availability date. For the full model-by-model list, see the GPT lineage table below or the ChatGPT models guide.

Last reviewed 29 September 2026. The history sections hold across releases. The DevDay 2026 and GPT-6 sections are dated and refreshed at each launch.

🗺️ OpenAI at a Glance: The Complete Evolution

🤖 What Is OpenAI?

OpenAI came to life in 2015 in San Francisco as a joint initiative of Elon Musk, Sam Altman, Greg Brockman, Wojciech Zaremba, Ilya Sutskever, and John Schulman. The mission was simple but wildly ambitious — develop safe and open AI tools to empower (rather than eradicate) people.

"OpenAI's mission is to ensure that artificial general intelligence (AGI) — by which we mean highly autonomous systems that outperform humans at most economically valuable work — benefits all of humanity."

OpenAI Charter(1)

What started as a nonprofit research lab has evolved into one of the most valuable companies on the planet. OpenAI exited 2025 at a roughly $20 billion annualized revenue run-rate, and by 2026 it served about a billion weekly ChatGPT users and had raised more than $160 billion, including a $122 billion round that closed in March 2026.

The company's product lineup has grown far beyond the original GPT models. ChatGPT is now a general-purpose AI assistant with browsing, code execution, image generation, and real-time voice conversations. Sora generates photorealistic video from text. The o-series reasoning models "think" before they answer. GPT-5, launched in August 2025, represents a paradigm shift in how AI models are built - prioritizing algorithmic efficiency over brute-force scale. And since September 2026, GPT-6 ships as a three-tier family, with dots, always-on agents, as the newest way to use it.

Who Owns OpenAI Today?

The OpenAI Foundation, a nonprofit, controls OpenAI Group PBC, the for-profit company that sells ChatGPT and the API, and Microsoft is the largest outside investor with roughly 27 percent. OpenAI completed this structure on October 28, 2025. It said the Foundation's equity was worth about $130 billion at the time, and Microsoft said its stake was about 27 percent on an as-converted diluted basis, counting all owners.

  OpenAI Foundation (nonprofit)
  Mission: AGI that benefits all of humanity
        |
        | controls
        v
  OpenAI Group PBC (for-profit public benefit corporation)
        |
        +-- Foundation equity ........ ~$130B at the Oct 2025 recapitalization
        +-- Microsoft ................ ~27% as-converted, diluted
        +-- Employees and investors .. the rest
        |
        +-- Sells: ChatGPT, the API, Codex, ChatGPT Work, dots

Sources: OpenAI's "Built to benefit everyone" post and Microsoft and OpenAI's joint partnership announcement, both October 28, 2025. Stake sizes change with every funding round, including the $122 billion round that closed in March 2026, so treat the percentages as the October 2025 snapshot.

What Does OpenAI Sell?

OpenAI sells five kinds of product: a consumer assistant (ChatGPT), developer models (the API), coding agents (Codex), work agents (ChatGPT Work and dots) and a build-your-own agent platform (the Agents API). Sora, its video product, shut down on April 26, 2026.

But the journey to get here was anything but smooth.

🥚 The History of OpenAI

The Early Days of Artificial Intelligence

So, what makes artificial intelligence?

It's simple. Add a dash of magic, sprinkle with fairy dust, and mix in a whole lot of caffeine-fueled all-nighters from computer scientists and engineers.

Ok, the answer is a tad more complicated than that.

Over the course of human history, we've been trying to animate objects and give them human-like qualities. But the first person who really pushed the needle was Alan Turing.

Turing's research in the early 1950s laid the foundation for modern computer science. AI was still a thing out of fiction novels, but it was enough to get more brilliant minds on board. That merry bunch included John McCarthy who coined the term "artificial intelligence" in 1956.

Two years later, McCarthy and his colleagues set up the Artificial Intelligence project at MIT. The future of AI research was starting to look bright, if only a little too optimistic.

After the initial interest, the AI bubble burst and funding dried up, mostly because of disappointing results and limited computing power. Some call this period the first "AI winter."

In the 1990s, advancements in machine learning and natural language processing (NLP) put artificial intelligence back in the spotlight. And a few publicity stunts helped it stay there.

In 1997, IBM's computer "Deep Blue" beat the World Chess Champion Garry Kasparov in a six-game match. That was AI's second triumph after a different computer named "The Oracle" had outmatched a Jeopardy! champion Ken Jennings a year earlier.

The early 2000s brought interesting developments like the explosion of big data, more refined algorithms, and increasing computing power. The door to advanced AI systems was wide open.

But the moment that actually kicked the door down came in 2012. Geoffrey Hinton and two students — Ilya Sutskever and Alex Krizhevsky — entered the ImageNet competition with a neural network called AlexNet, trained on two $500 gaming GPUs in Krizhevsky's bedroom. They obliterated the state-of-the-art, jumping accuracy from 74.3% to 84.7% on their first try. Twenty years of computer vision research were rendered obsolete with a single presentation. The recipe was absurdly simple: bigger neural networks + more data + GPU compute = better results. Within months, Google acquired Hinton's team for $44 million — buying not products but brains. The perceptron that Frank Rosenblatt built in 1957 had finally grown up, and it was hungry for compute.

OpenAI Joins the Party (2015–2017)

Over 65 years after Turing's breakthrough paper, the rapid expansion of AI systems raised a few eyebrows. It also made some people worried about the direction it was heading.

In 2015, former Y Combinator CEO Sam Altman and business "magnet" Elon Musk spearheaded an initiative for safe and open AI development. And that's where the history of OpenAI began.

Even before founding the company, Altman and Musk had expressed concerns about the risks and opportunities of AI technology, at some point calling it: "the greatest threat to humanity."

The founding group pledged $1 billion to the venture — a massive bet on a nonprofit with no products and no revenue. Other early backers included Peter Thiel, Reid Hoffman, and Amazon Web Services. The pitch was simple: if powerful AI was inevitable, it should be built by people who wanted it to be safe, open, and beneficial to everyone.

The company initially focused on developing artificial intelligence for video games and other applications. In 2016, it released its first tools, an open-source toolkit for reinforcement learning (RI) OpenAI Gym and Universe which was essentially a test-bed for training AI agents.

Pivot and Growth (2017–2019)

In the two years that followed, OpenAI focused on more general AI research and development.

In 2018, the company released a paper "Improving Language Understanding by Generative Pre-Training" — introducing the concept of a Generative Pre-trained Transformer (GPT).

In a nutshell, GPTs are neural networks — machine learning models inspired by the structure and function of the human brain — trained on a large dataset of human-generated text. It can perform many functions like generating and answering questions, among other things.

It can also write haikus about itself, like this one:

ChatGPT's mind vast,

Answers flow with ease,

AI's tongue at last.

Anyways... the OpenAI team put their money where their mouth was and developed GPT-1, their first language model "trained" on the BookCorpus containing over 7000 unpublished books.

The model eventually evolved into GPT-2, a more powerful version trained on 8 million web pages and containing 1.5 billion parameters (trained values that make text prediction possible).

At that point, the company pivoted from its lofty goal of "open" AI (pun not intended) and initially decided not to release GPT-2 to the public.

The (official) reason?

According to an OpenAI blog announcement, the team was afraid that GPT-2 could be used to write scam emails or generate fake news. And that made perfect sense. After all, great power comes with great responsibility (thanks, Peter Parker).

The company's pivot from its founding philosophy coincided with Elon Musk leaving the board of OpenAI in 2018. Musk cited a potential conflict of interest with Tesla's own AI work, but he later admitted he disagreed with the direction the company was taking. It was the beginning of a rift that would explode years later into lawsuits, hostile takeover bids, and one of Silicon Valley's most public feuds.

In 2019, OpenAI made another controversial decision to transition to a "capped-profit" organization and established OpenAI LP, "a hybrid of a for-profit and nonprofit." The reasoning was straightforward: developing frontier AI required billions of dollars, and nonprofits can't raise that kind of capital. But critics saw it as the first step away from the founding mission.

DALL·E, GPT-3, and ChatGPT (2021–)

In 2021, OpenAI released DALL-E, an AI using a similar architecture to GPT-2. Instead of generating text, DALL-E — a portmanteau of WALL-E and the Spanish surrealist artist Salvador Dalí — could create photorealistic images, seemingly out of thin air.

In 2022, OpenAI pushed the needle again with GPT-3. An iteration of the two previous models, GPT-3 was fed 45TB of text data that translated into 175B parameters. It was smarter, faster, and more terrifying than anything we had seen before.

To make that possible, Microsoft designed a supercomputer for OpenAI that included 285,000 CPU cores and 10,000 GPUs. It was also #5 in the ranking of Top500 supercomputers.

The success of GPT-3 spawned another monster in disguise.

On November 30, 2022, OpenAI released ChatGPT, a language model chatbot built on GPT-3.5, an upgraded member of the GPT-3 family. It was free and text-only at launch. Altman said it passed 1 million users in five days, and it was the fastest-growing consumer application in history, reaching 100 million users within just two months - a record that TikTok had taken nine months and Instagram two and a half years to achieve. OpenAI added its first paid tier, ChatGPT Plus at $20 per month, in February 2023.

One of the most jaw-dropping aspects of ChatGPT is its ability to understand context — the chatbot can generate answers and adjust them based on the conversation history. This means you can "train" ChatGPT within a conversation thread to get more accurate answers.

For many, the interaction with ChatGPT or DALL-E 2 — DALL-E's successor released the same month — was the first (conscious) and surreal contact with AI. It wasn't exactly love at first sight. But it made one thing clear — the future is AI, and it's already here.

(update) The GPT-4 Era (2023)

The growth of AI is like watching a toddler taking first wobbly steps. It's adorable and concerning, especially considering that GPT-4 arrived mere four months after ChatGPT.

According to OpenAI, GPT-4 was 40% more likely to provide accurate replies and 82% less likely to produce output that breaks content policy. The model could "understand" visual input, bumped the token limit to 8,192 tokens, and was OpenAI's first truly multimodal model.

GPT-4 quickly became the backbone of enterprise AI adoption, powering apps like Duolingo, Stripe, Khan Academy, and (a shameless plug) Taskade. Microsoft integrated it into Copilot across Office 365, GitHub, and Azure.

But the real disruption of 2023 wasn't a model launch — it was a boardroom coup (more on that below).

(update) The Great Exodus and the Race to Reason (2024)

If 2023 ended with the most dramatic CEO firing in Silicon Valley history, 2024 was defined by two things: departures and reasoning.

The departures came first. In February, AI researcher Andrej Karpathy left OpenAI for the second time to pursue personal projects. By mid-2024, co-founder and chief scientist Ilya Sutskever — the man widely considered the intellectual heart of the company — departed to start Safe Superintelligence Inc., a startup focused exclusively on building safe superintelligent AI. CTO Mira Murati resigned in September 2024, followed by VP of Research Barret Zoph, co-founder John Schulman (who joined Anthropic), and research lead Bob McGrew.

The exodus rattled the industry. Some called it a "brain drain," others a correction — safety-minded researchers leaving a company that was moving fast and breaking things.

Then came the reasoning revolution.

In May 2024, OpenAI launched GPT-4o ("omni"), a multimodal model that could process text, audio, and vision in real-time. It powered ChatGPT's voice mode with human-like intonation — the "Her" movie experience became real overnight. GPT-4o mini followed in July as the fastest, cheapest GPT-4 variant.

But the paradigm shift arrived in September with o1-preview, OpenAI's first "reasoning" model. Unlike traditional GPTs that generate responses in a single forward pass, o1 models "think step by step" before answering, dramatically improving performance on math, coding, and scientific reasoning. The full o1 launched in December 2024.

The year also brought three more milestones:

• ChatGPT Search (October 2024) — turning ChatGPT into a real-time search engine that could browse the web, directly challenging Google's core business.
• o3 (December 2024, preview) — the next-generation reasoning model that scored 87.5% on the ARC-AGI benchmark in high-compute mode, reigniting debates about the path to AGI.
• Sora Turbo (December 2024) — the public launch of OpenAI's text-to-video model, generating photorealistic clips up to 20 seconds. The implications for content creation, filmmaking, and marketing were staggering.

The Competitive Landscape (End of 2024):

OpenAI's valuation crossed $150 billion by the end of 2024. The race for AGI was no longer a distant dream — it was happening in real-time, with new model releases every few weeks from competitors around the world.

(update) From Stargate to GPT-5: The $500 Billion Bet (2025–2026)

If you thought 2024 was intense, 2025 made it look like a warm-up.

January 2025 kicked off with the announcement that shook the industry: the Stargate Project, a $500 billion joint venture between OpenAI, SoftBank, Oracle, and Abu Dhabi's MGX fund to build massive AI data center infrastructure across the United States. It was the largest single AI infrastructure commitment in history — and a signal that the path to AGI runs through power plants.

The same month, OpenAI launched Operator, an autonomous AI agent that could browse the web and complete tasks on behalf of users — booking flights, filling out forms, managing workflows. It was the company's first real step toward AI agents that don't just answer questions but take action.

Then came the DeepSeek shock. In late January 2025, Chinese AI lab DeepSeek released its R1 reasoning model, which matched o1's performance at a fraction of the training cost. The news wiped nearly $1 trillion off the Nasdaq in a single day — the largest single-day loss in stock market history. It challenged the assumption that only massive compute budgets could produce frontier AI, and it lit a fire under every research team in Silicon Valley.

February 2025 brought drama from an unlikely source. Elon Musk, who had been suing OpenAI since early 2024 over its departure from its nonprofit mission, made a jaw-dropping $97.4 billion bid to buy the company outright. The OpenAI board rejected it. The same month, OpenAI launched Deep Research, a feature that could spend minutes synthesizing comprehensive reports from across the web.

March 2025 saw OpenAI close a $40 billion Series F round at a $300 billion valuation — the largest private funding round in history, with SoftBank leading. Then GPT-4o's image generation capabilities went viral when users discovered they could generate art in the style of Studio Ghibli. The internet was flooded with Ghibli-fied photos, and the feature became one of ChatGPT's most viral moments since launch.

April–May 2025 brought rapid-fire product launches. The o3 and o4-mini reasoning models shipped in April, pushing the boundaries of mathematical and scientific reasoning. In May, OpenAI made two blockbuster acquisitions: Windsurf, a coding assistant platform, for $3 billion, and io (the Jony Ive hardware startup) for $6.4 billion. The Windsurf deal signaled OpenAI's ambition to own the developer tools market. The io acquisition hinted at a dedicated AI hardware device — a "post-smartphone" form factor designed by the same mind behind the iPhone.

August 7, 2025 — GPT-5 launched. And it was a mess.

The rollout was what Sam Altman later called "bumpy." GPT-5 had impressive capabilities — vastly better coding, scientific reasoning, and healthcare applications — but it also produced errors in areas where GPT-4 had been reliable. Altman acknowledged publicly that they had "totally screwed up some things" during the deployment. But the model itself represented a fundamental shift: GPT-5 used less training compute than its predecessor GPT-4.5 (codenamed Orion), which had been an internal disappointment — considered "too big" relative to its gains. The lesson was clear — algorithmic efficiency matters more than brute-force scale.

The same month, OpenAI launched ChatGPT Go, an $8/month tier designed to bring AI to price-sensitive markets worldwide.

September–October 2025 saw Sora 2 launch with dramatically improved video generation and the second annual DevDay conference where OpenAI announced apps inside ChatGPT, turning the chatbot into a full-fledged platform. On October 28, 2025, OpenAI completed its transformation into a Public Benefit Corporation (PBC), ending the hybrid nonprofit/capped-profit structure that had defined the company since 2019. Sam Altman received equity in the company for the first time.

By October, ChatGPT had reached 800 million weekly active users, a number Altman announced at TED.

November 2025 brought a quiet but symbolically massive change: OpenAI updated its mission statement, removing the word "safely" from its description of ensuring AGI benefits all of humanity. For an organization born out of AI safety concerns, the edit spoke volumes. Shopping features arrived the same month, with ChatGPT Shopping Research and later Instant Checkout turning the chatbot into a product discovery and purchasing tool.

December 2025 was an avalanche. GPT-5.2 shipped with improved reasoning. Disney invested $1 billion in OpenAI and signed an exclusive partnership to use Sora for film production. ChatGPT hit 900 million weekly active users. And OpenAI kicked off a $100 billion fundraise at an $830 billion valuation — which would make it the most valuable private company in history by a wide margin.

January 2026 saw OpenAI begin testing advertisements in ChatGPT — a move that surprised no one given the $8.5 billion in 2025 losses, but still drew criticism from users who feared a "Google-ification" of the product.

February 2026 brought two major moves. GPT-5.3-Codex launched as a specialized coding variant powering Windsurf and ChatGPT's code interpreter. And OpenAI hired Peter Steinberger, creator of OpenClaw — the open-source AI agent framework with 196,000+ GitHub stars — to lead next-generation personal AI agent development. OpenClaw transitioned to an independent foundation sponsored by OpenAI, keeping its open-source status while giving OpenAI access to its local-first, model-agnostic agent architecture. Sam Altman noted: "The future is going to be extremely multi-agent." OpenAI continues to expand internationally with offices in Tokyo, Singapore, Seoul, Paris, Brussels, and Mumbai.

March to September 2026 turned OpenAI's model line into a family. GPT-5.4 (March 5) brought a 1.05-million-token context window and native computer use, and a funding round closed on March 31 at $122 billion committed and an $852 billion valuation. GPT-5.5 followed on April 23. OpenAI shut down the Sora web and app experiences on April 26 and signed an amended, non-exclusive Microsoft deal on April 27. On June 8 it confidentially submitted a draft S-1, though Sam Altman later said an IPO would come "not 2026."

GPT-5.6 arrived in three tiers (Luna, Terra and Sol) as a limited preview on June 26 and generally on July 9, 2026, the same day OpenAI launched ChatGPT Work, an agent inside ChatGPT that produces finished sheets, slides, docs and Sites, built on Codex technology. A July 30 price cut took GPT-5.6 Luna down 80%, and by August OpenAI said a billion people used ChatGPT every week.

Then came the pacing summer. On August 18 OpenAI paused part of its reinforcement-learning work over cyber-critical capabilities. On September 3, 2026 it released GPT-6 Astra — "the world's most intelligent and aligned model," and the first it rated Critical in cybersecurity — at $10 per million input tokens and $50 per million output. On September 22, GPT-6 Sol ($2/$10) and GPT-6 Luna ($0.10/$0.50) brought Astra's training methods to everyday prices, the same morning Anthropic released Claude Opus 5.5. Our explainer on GPT and GPT-6 covers the three models, their benchmarks and which one to use.

Seven days later came DevDay. On September 29, 2026, OpenAI shipped GPT-6.1 Sol, which replaced GPT-6 Sol at the same $2 / $10 price, and launched dots, always-on agents powered by GPT-6 Astra. It also added an Ultrafast speed tier, a $500 Pro plan and computer use in its Agents API. The DevDay 2026 section below covers each launch, and our guide to always-on AI agents explains how the pattern works.

Complete Timeline of 2024–2026 Releases:

🧬 GPT Lineage: Every Release Date from GPT-1 to GPT-6.1 Sol

OpenAI has shipped a new flagship GPT model roughly every few months since GPT-4, and the cadence shortened to weeks in 2026: GPT-5.6 on July 9, GPT-6 Astra on September 3, GPT-6 Sol and Luna on September 22, and GPT-6.1 Sol on September 29. Prices below are OpenAI's API list prices per million tokens (input / output) at launch, from OpenAI's announcements and API changelog.

GPT-1, GPT-3 and GPT-3.5 rows follow OpenAI's public history. Later rows follow OpenAI's launch pages and API changelog, checked on 29 September 2026.

The pattern to read from the table: the flagship price stopped climbing and the middle tier collapsed. GPT-5.6 Sol launched at $5 / $30, fell to $4 / $20 in August, and reached $2 / $10 as GPT-6 Sol, then GPT-6.1 Sol kept that price with better results.

Output price fell from $30 to $10 in about ten weeks. A price per million tokens is not a bill, so here is the arithmetic for one job. The token counts are an illustrative assumption, not a measurement: 100,000 tokens in and 20,000 tokens out.

  Cost of one job = (input tokens / 1M x input price) + (output tokens / 1M x output price)
  Assumed job     = 100,000 tokens in, 20,000 tokens out
  GPT-6 Luna   (0.10 in / 0.50 out)   0.1 x $0.10 + 0.02 x $0.50  =  $0.01 + $0.01  =  $0.02
GPT-6.1 Sol  (2 in / 10 out)        0.1 x $2    + 0.02 x $10    =  $0.20 + $0.20  =  $0.40
GPT-6 Astra  (10 in / 50 out)       0.1 x $10   + 0.02 x $50    =  $1.00 + $1.00  =  $2.00
  Astra costs 5x Sol and 100x Luna for the same token counts.
Real jobs differ: a model that writes more tokens, or retries, costs more per task.

To see how a per-token price turns into a per-task bill, read AI cost per task, and for the tier logic behind Sol, Luna and Astra read What Is GPT?.

🆕 DevDay 2026: GPT-6.1 Sol, Dots, and the Always-On Agent Era

On September 29, 2026, OpenAI held DevDay 2026 in San Francisco and launched dots, always-on agents powered by GPT-6 Astra, together with GPT-6.1 Sol, a model that scores one point below Astra on the Artificial Analysis Intelligence Index at one-fifth of Astra's price. OpenAI called it its biggest DevDay yet, with more than 20 major announcements across ChatGPT, Codex, models and agents, and cited 1.2 billion weekly users.

Facts in this section were checked on 29 September 2026 against OpenAI's DevDay recap, its dots and GPT-6.1 Sol announcements, its help center and API changelog, and Artificial Analysis. Press-reported items are attributed to their outlet. OpenAI's benchmark charts are vendor claims.

The GPT-6 Family in One Table

The GPT-6 family has three tiers, Astra, Sol and Luna, with no Terra tier so far, and its output prices span 100x from $0.50 to $50 per million tokens. GPT-5.6 had Luna, Terra and Sol. GPT-6 dropped the middle tier and, one week after launch, upgraded Sol.

Luna costs one hundredth of Astra per output token. That spread is why routing matters more than picking one model, which is the argument in What Is GPT?.

GPT-6.1 Sol: Near-Astra Intelligence at One-Fifth of the Price

GPT-6.1 Sol replaced GPT-6 Sol after just seven days and, in Artificial Analysis's independent check published September 29, 2026, scores one point below GPT-6 Astra while costing $0.72 per task at maximum effort against Astra's $3.26. It costs $2 per million input tokens, $0.10 cached and $10 output, and runs in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu plans. It is not yet in Chat.

Three details make the launch worth reading closely:

• OpenAI's own chart. On Terminal-Bench Science at maximum effort, OpenAI reports $5.47 per task for GPT-6.1 Sol against $23.21 for Claude Opus 5.5 and $23.80 for Astra, while Astra keeps the top score. That is a vendor claim.
• More tokens per answer. Artificial Analysis found that GPT-6.1 Sol uses about 10 to 30 percent more output tokens than GPT-6 Sol, so cost per task, not price per token, is the number to compare.
• Safety rating. OpenAI treats GPT-6.1 Sol as Critical in cybersecurity and High in biology and chemistry under its Preparedness Framework, and it uses the Astra safeguards stack.

OpenAI says Sol Ultrafast, up to 8x faster in Codex, is coming soon. Our Codex pricing guide tracks how these models map to plans.

Dots: OpenAI's Always-On Agents

Dots are always-on agents in ChatGPT that OpenAI describes as "remarkably capable, always-on agents built to handle everything," powered by GPT-6 Astra, each with its own cloud computer and browser and a connection to more than 4,000 apps through plugins. A dot works toward your goals while you are away, and you reach it in ChatGPT on desktop, web and mobile, in Slack and Teams, and by voice in ChatGPT.

Regions matter. Pro access excludes the EEA, Switzerland and the UK at launch, while Business Premium is available in all supported ChatGPT regions. Your first dot is included in the plan at no extra cost, and you create it on desktop, because mobile creation is not supported at launch. OpenAI is also piloting specialist dots with their own identity and access to systems of record in enterprise pilots, and a Microsoft Agent 365 integration is in progress.

Read the diagram from left to right: research is read-only, action is gated twice, and you can see everything. OpenAI adds its own caveat: "Dots can still make mistakes, so always review consequential work." Dots are a consumer and team product, not an API. Developers get a similar architecture through the Agents API below. For the pattern behind dots, see always-on AI agents, the earlier persistent agents guide, and our notes on AI guardrails.

The Agents API, Ultrafast, and the New Pro Plans

OpenAI's Agents API entered public beta on September 10, 2026, and on September 29 it added computer use, where agents work in an OpenAI-hosted browser while your application handles website approvals and sign-in. OpenAI says it brings "that same harness and infrastructure that powers Codex" to developers and charges no extra fee beyond tokens and tools. You choose where the agent runs: an OpenAI-hosted sandbox, your own infrastructure, or a partner such as Cloudflare, Modal or Vercel.

That is OpenAI's agent ladder after DevDay: one model family feeding a personal agent, a task agent, a coding agent and a build-your-own API. Two more launches complete the picture. Ultrafast is a premium speed tier that OpenAI says runs up to 8x faster in Codex and up to 6x in the API, at six times the standard price for Astra ($60 in, $300 out per million tokens). The Decisions API, in limited preview, lets GPT-6 Luna answer with a fixed set of answers, which suits classifying, routing or picking an action.

Pro 500's 25x and Ultrafast come from OpenAI's DevDay recap and help center. The 5x and 10x multiples come from OpenAI's Codex lead on X, as reported by BGR. Existing Pro 200 subscribers keep the old allowance through October 29, 2026. Checked 29 September 2026.

The Pro change matters for anyone budgeting for Codex or ChatGPT Work: buying credits on Pro 100 or Pro 200 does not unlock Ultrafast at launch. See Codex pricing explained for how the plans compare.

The Agent-Product Graveyard: What OpenAI Retired

OpenAI retired or scheduled to retire five agent-related products in 2026, and three of them left within a year of launch, so the agent layer moves faster than the model layer. Every date below comes from OpenAI's release notes and API deprecations page.

If you build on a vendor's agent product, plan for its replacement. Model choice, agent memory and workflows are safer when they live in a layer you control. That is the design goal of Taskade: a workspace where AI agents keep persistent memory next to your projects, automations run them on triggers from connected apps such as a new email or a form response, and Auto picks among frontier models from top AI labs. Browse live community apps or start from Taskade Genesis to see it working.

💰 The Numbers Behind OpenAI

The financial story of OpenAI is as remarkable as its technology:

The gap between revenue and profitability tells the story of AI's economics. Training and running frontier models requires massive GPU clusters, and the Stargate Project's $500 billion commitment signals that compute costs will only grow. OpenAI projects $14 billion in losses for 2026.

Still, the growth trajectory is remarkable. Revenue roughly doubled in 2025, and the user base grew from 300 million to 900 million weekly active users in a single year. ChatGPT Plus ($20/month), the ChatGPT Pro tiers ($100, $200 and, since September 29, 2026, $500 per month), and enterprise API contracts drive the bulk of revenue, with ChatGPT Go ($8/month) opening up emerging markets.

Funding History:

The big question: can OpenAI grow revenue fast enough to justify an $852 billion valuation while burning billions per year? It's the same question Amazon investors faced in the early 2000s — and we know how that turned out.

🔎 Microsoft and the $13 Billion Partnership

Microsoft's relationship with OpenAI is one of the most consequential partnerships in tech history.

It started in 2019 with a $1 billion investment and a deal to make Microsoft Azure OpenAI's exclusive cloud provider. Then came $10 billion more in January 2023, followed by additional investments that brought Microsoft's total commitment to roughly $13 billion.

In return, Microsoft got to integrate OpenAI's models across its entire product stack — from Copilot in Office 365 and GitHub to Bing Chat and Azure AI services. The strategy was clear: use OpenAI's models to reinvent every Microsoft product for the AI era.

Today, we're launching an all new, AI-powered Bing search engine and Edge browser, available in preview now at Bing.com, to deliver better search, more complete answers, a new chat experience and the ability to generate content. We think of these tools as an AI copilot for the web.

Yusuf Mehdi, Corporate Vice President & Consumer Chief Marketing Officer(6)

But the partnership has grown complicated. As OpenAI's valuation soared past $500 billion and the company restructured into a PBC, questions arose about Microsoft's influence, profit-sharing arrangements, and potential conflicts. Microsoft now offers competing models alongside OpenAI's products, and OpenAI has quietly explored relationships with other cloud providers.

It's the kind of marriage where both sides benefit enormously but keep one eye on the exit.

🤯 The OpenAI Power Struggles

If OpenAI were a TV drama, no network would pick it up — the plot would seem too unrealistic.

The Five-Day CEO Crisis (November 2023)

On Friday, November 17, 2023, the OpenAI board fired Sam Altman as CEO. The announcement, delivered via a blog post, cited a loss of confidence in his ability to continue leading the company. No specific reasons were given.

What followed was five days of chaos:

• Day 1: Former Twitch CEO Emmett Shear was named interim CEO. President Greg Brockman resigned in protest.
• Day 2: Reports emerged that the board was reconsidering. OpenAI investors, led by Thrive Capital, pushed for Altman's return.
• Day 3: Microsoft CEO Satya Nadella announced that Altman and Brockman would join Microsoft to lead a new AI research lab. Nearly 500 of OpenAI's 770 employees signed a letter threatening to follow.
• Day 4: Behind-the-scenes negotiations intensified. The remaining board members faced the prospect of a company with no employees.
• Day 5: Altman returned as CEO with a new board that included Bret Taylor (former Salesforce co-CEO) as chair and Larry Summers.

The crisis exposed a fundamental tension at the heart of OpenAI: could a nonprofit board govern a company racing to build the most powerful technology in human history while managing billions in commercial partnerships?

The answer, apparently, was no.

The Musk Saga

Elon Musk's relationship with OpenAI reads like a breakup that never ends. After co-founding the company in 2015 and leaving the board in 2018, Musk grew increasingly vocal about what he saw as OpenAI's betrayal of its open-source, nonprofit roots.

In February 2024, Musk filed a lawsuit against OpenAI and Altman, alleging they had broken their founding agreement by prioritizing profit over public benefit. The suit was dropped, refiled, and became a running legal battle.

Then in February 2025, Musk made his dramatic $97.4 billion bid to buy OpenAI outright, partnering with a group of investors. The board rejected the offer. The feud became personal, playing out in real-time on social media for millions of followers — even as Musk's own company xAI (and its Grok chatbot) competed directly with ChatGPT.

The Departures and the Mission Drift

Beyond the headline drama, a quieter but more consequential shift was happening inside OpenAI. In summer 2023, Sutskever had paused his AI development research to launch Super Alignment — a project to ensure superintelligent AI could be controlled, with a 4-year timeline. Altman publicly committed 20% of OpenAI's compute to the effort. But the compute never came in the promised quantities. ChatGPT's growth, enterprise deals, and maintaining the lead over Google took priority. Safety kept getting pushed to the next quarter. The departure of safety-focused researchers throughout 2024 — Sutskever, Murati, Schulman, Zoph — wasn't just about talent. It signaled a cultural transformation.

The OpenAI of 2015 was a nonprofit research lab obsessed with safe AI development. The OpenAI of 2025 was a $500 billion PBC with 900 million users, advertising revenue, shopping features, and a mission statement that no longer contained the word "safely."

Was this mission drift? Or pragmatic evolution? Supporters argue that you can't influence the future of AI from the sidelines — you need revenue, users, and scale to matter. Critics counter that OpenAI became exactly the kind of company it was founded to prevent: a commercially driven AI giant with an insatiable appetite for capital.

The truth, as usual, is probably somewhere in the middle.

The Corporate Structure: From Nonprofit to $500B PBC

OpenAI's corporate structure has been one of the most debated topics in tech. Here's how it evolved:

The PBC structure means OpenAI must legally balance shareholder returns with its public benefit mission. In theory, this prevents a pure profit-maximizing approach. In practice, critics argue that the mission language is vague enough to justify almost anything — especially after the word "safely" was removed.

The original nonprofit entity still exists and holds a stake in the PBC, but its governing power has been dramatically reduced. For all practical purposes, OpenAI is now a for-profit company with a mission statement attached.

🎙️ Inside the Founder's Mind: Sam Altman from GPT-5 to DevDay 2026

At DevDay 2026, Sam Altman described dots as "the way I've always wanted to work with AI," and the agent-first framing continues themes he set out during the GPT-5 era. Quotes in the next block come from CNBC's DevDay live blog on September 29, 2026, unless noted.

At DevDay 2026: Agents, Plugins and Hardware

• On dots: "It's the way I've always wanted to work with AI." CFO Sarah Friar described dots as "the agent that can tap you on the shoulder."
• On plugins: Altman said plugins can be "essentially entire applications that feel native to ChatGPT."
• On hardware: Asked about a device, he said there could be "something that's worth waiting for" and that he is "not worried about being first." No device appeared at DevDay.
• On agent safety: He called Nvidia's new agent safety platform "a good thing" but "not a full solution."
• His closing line: "AI should be about giving people more power over their own lives."

Not everything worked on stage. Simon Willison's DevDay live blog records that several live demos, including dot voice and CLI voice, failed, and OpenAI's Tibo Sottiaux said the company would keep doing live demos.

The GPT-5 Era Interview

The longer interview below is from the GPT-5 era. It stays here because its ideas about agents, energy and compute explain the DevDay 2026 launches. In a wide-ranging conversation on the Huge Conversations podcast, Altman offered a rare, candid look at where OpenAI was heading, and what kept him up at night.

On GPT-5's Capabilities (and Its Bumpy Launch)

Altman acknowledged that GPT-5's rollout was far from perfect — the model excelled at coding and scientific reasoning but stumbled in areas where GPT-4 had been reliable. He described asking GPT-5 to build a Snake game, reminiscing about writing the same game on a TI-83 calculator as a kid. The fact that an AI could now do what took him hours of teenage tinkering was, for Altman, both thrilling and vertigo-inducing.

What's less discussed is the engineering decision behind GPT-5: it used less training compute than GPT-4.5 (codenamed Orion), which had been considered "too big" internally. Orion was a cautionary tale — the team had scaled up training compute dramatically and gotten diminishing returns. So for GPT-5, they pivoted from brute-force scaling to focusing on algorithmic breakthroughs.

Altman described three paradigm shifts in AI development: the original GPT paradigm (predict the next token, scale the data), the reasoning paradigm (o1/o3 chain-of-thought), and a new approach still emerging that combines both with novel architectures. The implication is that the next leap won't come from bigger models — it'll come from smarter ones.

On Superintelligence

When pressed on when we might see superintelligent AI, Altman offered his working definition: a system that is both better than OpenAI's research team at AI research and better than him at running OpenAI. By that standard, he believes we're within striking distance.

He predicted that within one to two years, AI will make a significant scientific discovery — not just assist researchers, but independently identify and verify something new about the natural world. And he offered a provocative thought: a kid born today will never be smarter than AI.

On Energy, Compute, and the Infrastructure Race

The path to superintelligence, Altman argued, runs through power plants. Energy is the limiting factor, not algorithms. He described the need for gigawatt-scale data centers — facilities that consume as much electricity as a small city. The Stargate Project's $500 billion price tag starts to make sense in this context.

Altman has spoken publicly about OpenAI's interest in nuclear energy, and he sees the transition to clean energy as both an AI requirement and a societal benefit — AI forces the construction of energy infrastructure that the world needs regardless.

On the One-Person Billion-Dollar Company

One of Altman's most-quoted predictions: within the next few years, we'll see the first one-person billion-dollar company. AI agents will handle engineering, sales, customer support, and operations, leaving a single founder to set the vision and make strategic decisions. It's a radical reimagining of what entrepreneurship could become — and a challenge to the entire venture capital model built on the assumption that scaling requires headcount.

On Health, Medicine, and the Social Contract

Altman was most animated when discussing healthcare. GPT-5, he said, is significantly better at medical reasoning than any previous model — not just as a diagnostic aid but as a partner that can synthesize patient data, research papers, and treatment protocols in real-time.

More broadly, Altman argued that AI will require rethinking the social contract. If AI can do most economically valuable work, how do we distribute the benefits? His answer: AI compute as a public resource, distributed like a form of universal basic intelligence. The transistor, he noted, was one of the most important inventions of the 20th century — but its true impact came not from the device itself but from the ecosystem it enabled. AI, he believes, will follow the same trajectory.

On Facts vs. Truth (The Jensen Huang Question)

In one of the interview's most philosophical moments, the conversation turned to a question posed by Nvidia CEO Jensen Huang: what's the difference between facts and truth? Altman's response revealed how he thinks about AI alignment at a deeper level — the challenge isn't just getting AI to produce accurate facts, but to understand nuance, context, and the messy human reality that sits between data points.

It's the kind of question that separates chatbot-level AI from something approaching genuine intelligence. And for Altman, it's the frontier that GPT-5 begins to explore — not just processing information, but engaging with it in ways that feel meaningfully different from pattern matching.

On Synthetic Data and the Training Data Wall

One of the most technically interesting parts of the interview dealt with synthetic data — AI-generated training data used to train the next generation of models. The concern in the industry has been that the internet's supply of high-quality human-generated text is finite, creating a "data wall" that limits future model improvements.

Altman was bullish on synthetic data, arguing that carefully constructed synthetic datasets can be as valuable as organic data — if not more so. The reasoning models (o1, o3) already rely heavily on synthetic reasoning traces for training. This shift has massive implications: it means OpenAI's future models may improve primarily by learning from their own outputs, a recursive loop that some researchers find promising and others find deeply concerning.

🤔 So, What Makes OpenAI Different?

Cross-Pollination of Ideas

OpenAI is uniquely positioned in the industry. It has a set of viable products, the best talent in the game, and great publicity, spun in large part by ChatGPT's explosive adoption.

And it's much easier to make friends and share ideas when you have something to show.

The company's partnerships now span every major industry. Microsoft integrates GPT models across its entire product stack. Disney invested $1 billion and is using Sora for film production. And platforms like Taskade leverage OpenAI's models alongside Anthropic and open-weight providers, giving users access to 15+ frontier models in a single workspace.

By engaging with partners across different industries, OpenAI can expand into new areas like healthcare, transportation, education, and finance. Plus, there's the hard-to-turn-down opportunity to access a vast amount of real-world customer data to train its AI models.

Responsible AI Development (A Complicated Legacy)

Ethics in AI is like trying to teach a robot to tell jokes, it's always a little bit off.

Said AI in a perfectly composed tone.

Ok, let's get serious for a moment.

OpenAI's relationship with safety has evolved dramatically. The company was literally founded to prevent dangerous AI, and its early research on adversarial attacks, fairness, and alignment was groundbreaking.

But the company's actions have increasingly diverged from its rhetoric. The November 2025 mission statement edit, the departure of safety-focused researchers, and the breakneck pace of releases have fueled criticism that OpenAI prioritizes speed over caution. Tools like Sora, the image generator, and autonomous agents like Operator all raise thorny questions about misuse, deepfakes, and accountability.

The counterargument is that deploying models at scale is itself a form of safety testing — you learn more from 900 million users than from any red-teaming exercise. Whether that reasoning justifies the pace is one of the defining debates in AI.

The Platform Play

What sets OpenAI apart in 2026 isn't just model quality — it's the ecosystem. ChatGPT has evolved from a chatbot into a platform:

• ChatGPT Search — a real-time search engine competing with Google
• Shopping Research + Instant Checkout — product discovery and purchasing
• Sora: video generation; the Sora web and app experiences shut down on April 26, 2026
• Dots and ChatGPT Work: always-on agents and finished-document agents, built on the GPT-6 family
• Operator: the January 2025 web-browsing agent that started OpenAI's agent line
• Apps in ChatGPT — third-party applications inside the chat interface
• API + Windsurf — developer tools for building on OpenAI's models

This platform strategy mirrors what Apple did with the iPhone: build the best product, then let others build on top of it. It's a bet that ChatGPT will become the default interface for how people interact with AI — and potentially the internet itself.

⚡️ Potential Benefits of OpenAI

It's hard to predict how the development of ChatGPT, GPT-5, and other tools will shape the world. But the impact is already visible across industries.

AI systems are already reviewing X-ray and CT scans, analyzing patient records, generating legal documents, writing production code, and tutoring students in real-time. GPT-5's improvements in medical reasoning suggest healthcare could be one of the biggest beneficiaries.

Sam Altman's prediction of one-person billion-dollar companies points to a world where AI dramatically lowers the barrier to entrepreneurship. Tools like Taskade AI already let individuals and small teams automate workflows, generate content, manage projects, and build AI agents — capabilities that previously required entire departments.

Whether we like it or not, the future of AI is in the hands of companies like OpenAI which will play a critical role in shaping the landscape and defining what "beneficial" use means.

Beyond Chat: When AI Builds Your Tools

ChatGPT proved that AI understands language. Taskade Genesis takes the next step — AI that does not just answer questions but builds live applications. Where ChatGPT gives you text, Taskade Genesis gives you a deployed dashboard, CRM, or client portal with AI agents that keep working after you close the tab.

Taskade gives you 15+ frontier models from OpenAI, Anthropic, and open-weight providers, with Auto mode picking a model for each request. The difference is what happens after the prompt: ChatGPT returns a response. Taskade Genesis returns a running system with Workspace DNA — Memory feeds Intelligence, Intelligence triggers Execution, Execution creates Memory.

Try OpenAI models inside Taskade Genesis →

OpenAI Codex and the Agentic Turn (2026)

OpenAI's biggest 2026 shift wasn't a new chatbot - it was agents that build software. OpenAI Codex, reborn in 2025 as an autonomous coding system, crossed 5 million weekly users by June 2026, and roughly one in five are non-developers. On June 2, 2026 OpenAI launched Codex Sites: describe a dashboard or internal tool in plain language and Codex builds, deploys, and hosts it. Sites started on ChatGPT Business and Enterprise and reached Plus and Pro in public beta on July 9, 2026. Launch coverage from VentureBeat and 9to5Mac describes sharing as a link for people inside your workspace.

That's the same prompt-to-app idea Taskade Genesis ships for everyone. Where Codex Sites shares an app inside a workspace, Taskade Genesis publishes it to the open web and turns the build into a living app you own - even on a free plan (up to 3 live apps on a Taskade-hosted public URL). Business plans add custom domains and client logins. See the full OpenAI Codex history and Codex Sites breakdown.

👉 How to Get Started with ChatGPT

ChatGPT is available at https://chatgpt.com with a free tier, Go ($8/month), Plus ($20/month), and Pro plans at $100, $200 and $500 per month.

Since August 2026, GPT-5.6 Luna is the default on the Free and Go tiers, and Free and Go users can use GPT-6 Luna in the desktop app. Plus adds GPT-6 Astra in ChatGPT Work and Codex, and Pro includes GPT-6 Astra in Chat as well. Since September 29, 2026, GPT-6.1 Sol is available in ChatGPT Work and Codex on Plus, Pro, Business, Enterprise and Edu, though not yet in Chat, and dots are rolling out to Pro and Business Premium.

You can ask questions, write code, generate images, browse the web, analyze documents, and complete tasks with Operator. ChatGPT also supports custom GPTs — specialized chatbots that anyone can build without coding. Need inspiration? Browse hundreds of ready-made AI prompts to see what ChatGPT and other models can do.

💡 Pro Tip: Run out of ideas or not sure where to start? Taskade AI includes a gallery of prompt templates you can use in Taskade or in other prompt-based AI tools out there!

🧬 Where Taskade Genesis Fits: Build on Top of Any Model

The model is one layer of a working system. Taskade Genesis turns one prompt into a live, hosted app with its own data, AI agents and automations, and the model underneath can change without your app changing with it. OpenAI's history shows why that matters: Pulse, the Atlas browser, Agent Builder and Custom GPTs each had a short life. Projects remember, agents think and automations execute. That is Workspace DNA: Memory + Intelligence + Execution, in one workspace.

Here is how OpenAI's September 2026 agent ladder maps to what Taskade Genesis offers for the same jobs.

                ▲  MEMORY  (Projects)
                   Your data, docs and history.
                   Databases that feed every agent.
                  /                              \
     writes back /                                \ feeds
                /                                  \
   ● EXECUTION                                  ■ INTELLIGENCE
     (Automations)                                (AI agents)
     Triggers, actions, schedules,                Reason over Memory,
     100+ integrations.                           use tools, decide.
                \                                  /
                 \____ triggers the next action ___/
   Execution creates Memory. Memory feeds Intelligence. Intelligence triggers Execution.
One prompt in Taskade Genesis builds all three, plus the app people see.

The agents here read the web and act through your connected apps. They do not drive a browser or a computer. That keeps every step something you configured and can audit.

What You Can Build With Taskade Genesis

One prompt builds the app. These examples come from the Taskade Genesis guides.

A Finance Tracker built with Taskade Genesis. Clone it free, then change it to track any company's rounds and models.

Taskade lets you pick a model for each agent when a job needs it, so a change in the OpenAI lineup, such as the move from GPT-5.6 to GPT-6, does not force a rebuild. Teams of agents work the same way. Describe the team in plain English, and Taskade EVE creates the specialists, which share one workspace memory. Read what are AI agents, multi-agent systems, AI subagents vs agent teams and autonomous task management for the patterns, or best OpenClaw alternatives for the always-on landscape. The newsletter posts Agents Think, Automations Execute, Set It Once and Memory, Intelligence and Execution as one multi-agent workspace show these in action.

Pricing is transparent. Every price below is billed annually, with monthly billing at $20, $50, $200 and $500 for Pro, Business, Max and Enterprise.

Comparing tools? Read the fair, side-by-side pages for Codex, Devin, Emergent, Manus and Base44, and see current pricing. Build your first app →

🚀 Quo Vadis, OpenAI?

The next chapter of OpenAI's story will be defined by three questions:

When will AGI arrive? Altman has hinted that GPT-5 and its successors are approaching the threshold. His own definition — an AI better than his research team at research and better than him at management — suggests it could happen within a few years. The ARC-AGI benchmark scores and rapid improvements in reasoning models suggest the timeline is closer than many expect.

Can OpenAI afford the race? With $8.5 billion in 2025 losses, $14 billion projected for 2026, and profitability not expected until 2029, OpenAI is betting that the AI market will be so large that today's losses won't matter. The $100 billion fundraise at $830 billion valuation is both a vote of confidence and a reminder of how much capital the AI race devours.

Will agents become the interface? Dots, ChatGPT Work and the Agents API point one way: OpenAI wants agents that keep working when you close the tab. The open questions are trust and control. OpenAI's own design keeps password changes with the human and adds Custom Rules and Auto-review, and it says dots "can still make mistakes." Read always-on AI agents for how the pattern works and how to run it on triggers you control.

Will the mission survive? From "ensure that artificial general intelligence safely benefits all of humanity" to a PBC with ads, shopping, and the word "safely" quietly removed — the transformation is real. OpenAI's challenge isn't just building AGI. It's building it in a way that the company's founders would still recognize.

One thing is clear — whether we're ready or not, we're heading towards a future where artificial intelligence will become a constant in our personal and professional lives. And OpenAI, for better or worse, is leading the charge.

🐑 Before you go... Looking for an AI tool that can seamlessly integrate with your task and project management workflow? Taskade AI has you covered!

• 💬 AI Chat: Got a question about your project? The AI Chat will give you all the answers and help your team stay on the same page. Decision-making has never been easier.

• 🤖 AI Agents: Tired of repetitive tasks? Let Taskade's AI Agents handle them. Agents work in the background, help your team save time, and let you focus on the big picture.

• ✏️ AI Assistant: Planning and organizing? You can use the AI Assistant to brainstorm ideas, generate content, organize tasks, and ensure you always know what to do next.

• 🔄 Workflow Generator: Starting a project can be a puzzle. Describe your project and the Workflow Generator will set things up for you, giving you a clear path to follow.

Want to give Taskade AI a try? Create a free account and start today! 👈

🔗 Related Reading

• Always-On AI Agents: How Dots and Persistent Agents Work - The pattern behind OpenAI's September 2026 launch

• What Is GPT? GPT vs LLM vs ChatGPT and the GPT-6 Tiers - Astra, Sol and Luna explained

• ChatGPT Models Explained: Every Version - Release-by-release catalog

• OpenAI Codex Pricing: Every Plan - How the Pro tiers and credits compare

• AI Cost per Task - Why price per token misleads

• What Are AI Claws? Persistent Autonomous Agents - The agents that stay on

• What Are AI Agents? - The definition behind dots and the Agents API

• Multi-Agent Systems - How teams of agents divide the work

• AI Subagents vs Agent Teams - Two ways to split a job across agents

• Autonomous Task Management - Agents that plan, act and check their own work

• Best OpenClaw Alternatives - The open-source always-on agent and its rivals

• What Is OpenAI Codex? From 2021 Code Model to Codex Sites — OpenAI's coding agent, deprecated then reborn

• What Is Anthropic? History of Claude AI - The safety-first lab behind Claude, OpenAI's closest rival

• What Is GitHub? Complete History — Octocat to Copilot, and the $7.5B Microsoft era OpenAI's Codex helped build

• What Is NVIDIA? Jensen Huang & the AI Revolution — The hardware powering GPT

• What Is Agentic Engineering? — From chatbots to autonomous AI agents

• The History of Computing — From binary to AI agents

• Claude Alternatives: 12 Best AI Assistants Like Claude 2026

🔗 Resources

1. https://openai.com/charter/

2. https://en.wikipedia.org/wiki/Bombe#/media/File:Bletchley_Park_Bombe4.jpg

3. https://www.kasparov.com/timeline-event/deep-blue/

5. https://openai.com/blog/chatgpt

6. https://blogs.microsoft.com/blog/2023/02/07/reinventing-search-with-a-new-ai-powered-microsoft-bing-and-edge-your-copilot-for-the-web/

8. https://openai.com/index/devday-2026-recap/

9. https://openai.com/index/introducing-dots/

10. https://openai.com/index/introducing-gpt-6-1-sol/

11. https://artificialanalysis.ai/articles/gpt-6-1-sol-replaces-gpt-6-sol-after-just-7-days-with-near-astra-intelligence

12. https://openai.com/index/introducing-the-agents-api/

13. https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers

14. https://www.cnbc.com/2026/09/29/openai-devday-2026-live-updates.html

💬 Frequently Asked Questions About OpenAI