AI news
AI model reviews
AI tools
How-to guides
Our staffAboutContact

OpenAI review 2026: features, pricing, who owns it, pros and cons

OpenAI makes ChatGPT, GPT-6 and Codex. We weigh the models, free and paid plans, API pricing, ownership and the Agent Builder shutdown as of late October 2026.

OpenAI (ChatGPT and API platform) image from its official website

Our verdict on OpenAI (ChatGPT and API platform)

4.5/ 5
4.5
Features
9.5
Ease of use
8.5
Value
8
Performance
8.5

OpenAI remains the broadest and most accessible AI platform in late 2026, with a useful free ChatGPT plan, cheaper GPT-6 models and a deep developer API. Everyday users, professionals and developers should use it; skip it if you need dependable human support, video generation or stable usage limits.

Pricing ChatGPT Free; Go $8/mo; Plus $20/mo; Pro $100, $200 or $500/mo; Business $25/seat/mo ($20 annual) or Premium $125 ($100 annual); Enterprise custom. API from $0.10/$0.50 per 1M tokens (GPT-6 Luna) to $10/$50 (GPT-6 Astra)

Visit OpenAI (ChatGPT and API platform)

I wanted to answer one question before writing a single word of this OpenAI review: if you type "open ai" into a search bar, what are you actually signing up for? A chatbot? A developer platform? A company that might be worth more than a trillion dollars? The honest answer is all three, and that's exactly why OpenAI is so easy to use and so confusing to pin down.

Here's the quick version. OpenAI is the company behind ChatGPT, the GPT-6 family of models, the Codex coding agent and the API that thousands of other apps are built on. You can use ChatGPT for free, the paid plans start at $8 a month, and in late October 2026 it's still the broadest AI product you can sign up for. It's also one of the most complained-about.

A quick confession before we go further: nobody on our team ran a hands-on test for this piece, so you won't find stopwatch timings or screenshots of my chats here. I'd rather tell you that than fake it. We reviewed OpenAI as it stands in late October 2026 using OpenAI's own announcements, pricing breakdowns, independent benchmarks, other reviewers' hands-on tests and a stack of user reviews. When someone else did the testing, I'll name them.

What is OpenAI?

OpenAI is an American AI research and product company. It started in 2015 as a nonprofit lab with a mission to make sure artificial general intelligence benefits everyone, and it's since turned into one of the biggest consumer tech businesses on the planet. If you've ever wondered "what is open ai" versus "what is ChatGPT", think of it this way: OpenAI is the company, ChatGPT is its best-known app, and GPT-6 is the brain inside.

The scale is hard to overstate. OpenAI says ChatGPT now serves more than 1.2 billion people each week. Reuters reported in late September that annualised recurring revenue was approaching $70 billion, though OpenAI hasn't confirmed that figure.

Who owns OpenAI?

This is the question people ask most, and the answer is a bit tangled. In October 2025, OpenAI finished a restructuring that turned its for-profit arm into a public benefit corporation still controlled by the nonprofit, now called the OpenAI Foundation. As part of that deal, Microsoft took roughly a 27% stake in the for-profit company, valued at $135 billion at the time, and kept access to OpenAI's technology until at least 2030.

So, did Microsoft buy OpenAI? No. Microsoft is the largest outside shareholder, but the Foundation keeps overall control. Other investors have piled in since: OpenAI raised $122 billion in March 2026 at an $852 billion post-money valuation, and Bloomberg and Reuters reported in September that it's in early talks to raise at least $30 billion more at around $1.4 trillion. Those later rounds will have shifted the exact percentages, and OpenAI hasn't published an updated cap table. It isn't publicly traded either: CEO Sam Altman has ruled out an IPO in 2026, according to the same report.

What does OpenAI do?

In practice, OpenAI makes three kinds of things:

  • Consumer apps: ChatGPT on web, desktop, iOS and Android, plus agent features like Codex and the Work agent inside it.
  • Models: the GPT-6 family (Astra, Sol and Luna), along with older GPT-5.6 models that are still on sale.
  • A developer platform: the OpenAI API, where businesses pay per token to build these models into their own products.

One product you can't use any more is Sora. OpenAI closed the Sora video app in April and switched off its API in September, which we covered in our Sora AI review.

OpenAI careers and jobs

If you landed here searching for "open ai careers" or "open ai jobs", you're not alone. The Financial Times reported in March that OpenAI had about 4,500 employees and planned to reach roughly 8,000 by the end of 2026, hiring across product, engineering, research, sales and "technical ambassadorship" roles. OpenAI hasn't confirmed those numbers. Open roles are listed on OpenAI's own website, and if you're thinking about the wider job market, our piece on AI jobs is a good next stop.

Key features

ChatGPT with GPT-6 and Intelligent UI

On October 7, OpenAI started rolling GPT-6 into ChatGPT. Paid plans (Plus, Pro, Business and Enterprise) got GPT-6 Sol first, and Free and Go users got the lighter GPT-6 Luna the next day. The headline change is Intelligent UI: instead of a wall of text, answers can arrive as charts, forms, tappable buttons and small interactive tools, with plain text used when that makes more sense. In OpenAI's words, ChatGPT "can combine UI, data, and actions" rather than squeezing every question into the same chat format.

Does it work? Mostly, from what reviewers and users report, but not for everyone. A widely read thread on OpenAI's own forum complains about cluttered, automatically generated UI elements that makes the product feel slower. If you want the deep dive on this update, read our ChatGPT GPT-6 review.

OpenAI models: Astra, Sol and Luna

The open ai models lineup now has three tiers, and the names finally make some sense:

  • GPT-6 Astra launched in early September and remains OpenAI's "highest-capability model", aimed at long agentic work like research, coding and computer use.
  • GPT-6 Sol is the mainstream workhorse. An updated GPT-6.1 Sol arrived at DevDay for coding, computer use and professional tasks, and OpenAI says it "approaches GPT-6 Astra on several evaluations".
  • GPT-6 Luna is the small, fast and very cheap model behind the free tier.

All three have a context window of about 1.05 million tokens with up to 128,000 output tokens, according to ComputingForGeeks. That's a lot of room for long documents and big codebases. If the jargon is getting thick, our explainer on what an AI model is will help.

Codex and the Work agent

Codex, OpenAI's coding agent, is included from the Plus plan up, alongside the Work agent and a site-building tool called Sites, according to The AI Career Lab. At DevDay, Codex gained cloud environments so you can kick off a task from your phone and let it run remotely, voice input and an /agents view in the CLI, automated pull request review for GitHub and GitLab, and Codex Security Cloud, which scans repositories and prepares fixes.

The OpenAI platform and API

For developers, the open ai platform is where the real breadth lives. DevDay added an Agents API with computer use, letting apps operate software through its graphical interface, plus multi-agent features borrowed from Codex, tool search and context compaction. There's also a Decisions API in limited preview that uses Luna to pick from a fixed set of answers, which is handy for classification and routing. On the ChatGPT side, developers can now build plugins with sidebars, panels and custom file viewers, and OpenAI previewed Dots (persistent agents for ongoing tasks) and ChatGPT Space, a shared workspace for teams and agents.

What happened to OpenAI Agent Builder?

Plenty of people still search for "open ai agent builder", so here's the news you need. Agent Builder launched in October 2025 as part of AgentKit, a visual canvas for composing logic with drag-and-drop nodes. It never left beta. In June 2026, OpenAI announced it's winding down Agent Builder and Evals, and both disappear from the platform on November 30, 2026. OpenAI recommends the Agents SDK if you build in code and Workspace Agents in ChatGPT if you'd rather describe a workflow in plain English. If you have workflows in Agent Builder, start moving them now.

OpenAI pricing and plans

OpenAI pricing splits into two worlds: ChatGPT subscriptions for people and teams, and pay-as-you-go API rates for developers. Here are the ChatGPT plans, based on The AI Career Lab's breakdown updated October 11, 2026 (checked October 2026):

PlanPriceWhat you get
Free$0GPT-6 Luna, unlimited text chats within abuse limits, capped uploads and tools, limited image generation
Go$8/monthAbout 10x Free's messages, uploads and images; may include ads
Plus$20/monthGPT-6 Sol, Codex, Work agent, Sites, no ads
Pro$100, $200 or $500/monthFrom 5x Plus usage; the $500 tier adds an Ultrafast mode
Business$25/seat/month ($20 annual); Premium $125 ($100 annual)Admin controls, SSO; Premium adds 5x usage and full GPT-6 Astra access
EnterpriseCustomCompliance controls, SCIM, data residency

One wrinkle: the $200 Pro tier reopened to new sign-ups on September 29 with a lower allowance than before, which annoyed some long-time subscribers.

Is OpenAI free? What the free plan really includes

Yes, ChatGPT is free to use, and the free plan is better than it used to be. You get GPT-6 Luna, Intelligent UI and what The AI Career Lab describes as "unlimited text chats", with guardrails against abuse. The limits bite elsewhere: uploads and tools are capped, and image generation is limited and slower, around two or three images a day in their observation. Codex, the Work agent and the stronger Sol model all need Plus or higher. If images are your main goal, our roundup of the best AI image generators has free options with more generous limits.

OpenAI API pricing

API costs are per million tokens. These are the headline rates for the GPT-6 family (checked October 2026):

ModelInputOutput
GPT-6 Astra$10.00$50.00
GPT-6 Sol / 6.1 Sol$2.00$10.00
GPT-6 Luna$0.10$0.50

Astra's rates come from eesel's summary of OpenAI's pricing page, which also notes Batch and Flex at half price, a Fast mode at double, and higher rates for requests over 272,000 input tokens. Sol and Luna prices were reported by TechEconomy, and cached input for GPT-6.1 Sol costs $0.10 per million tokens, according to InfoQ. Sol costing a fifth of Astra is a big deal: for most apps, it's the sensible default.

Performance and independent testing

Benchmarks first, with the usual warning that the vendor's own charts deserve a pinch of salt. On OpenAI's launch chart, as compiled by ComputingForGeeks, GPT-6 Astra scored 73.5% on OSWorld 2.0 (computer use) and 74.1% on DeepSWE 1.1 (coding), slightly ahead of Anthropic's older Claude Opus 5. Interestingly, GPT-6 Sol's best DeepSWE and OSWorld scores were a bit below the previous GPT-5.6 Sol. Its real improvement was cost.

The independent numbers tell a more mixed story. Artificial Analysis, as reported by ComputingForGeeks, found:

  • It's much cheaper to run. Sol's cost per task on the Intelligence Index fell to $1.06 from $1.99 for GPT-5.6 Sol, and Luna's to $0.07.
  • It makes things up less. Sol's hallucination rate on AA-Omniscience dropped from 92% to 60%, partly because it now declines to answer more often.
  • Some work quality slipped. Both models regressed on GDPval-AA, a test of real-world professional tasks, by about 100 Elo for Sol.

Against Anthropic's current flagship, it's close. In OpenAI's GPT-6.1 Sol chart, Claude Opus 5.5 edged GPT-6 Astra on AutomationBench (42.5% vs 41.4%) but trailed it on Terminal-Bench Science. My read: OpenAI and Anthropic are trading punches at the top, and you won't feel a dramatic gap either way for everyday work. Our Claude AI review covers the other side of that fight.

For hands-on coding tests, two YouTube creators are worth your time. AICodeKing put GPT-6.1 Sol through all eight KingBench 3 tasks in Codex, and AI Coding Daily ran GPT-6 Sol and Luna on 144 coding prompts. Both are useful if you want to see the models work rather than read a score.

Then there's the gap between ratings and reviews. The ChatGPT Android app sits at 4.8 stars from about 65.9 million ratings with over a billion installs, while ChatGPT's Trustpilot page shows a 1.6 out of 5 from 3,796 reviews, 74% of them one star. I read the 20 most recent Trustpilot reviews we could access (late August to early September) and the GPT-6 complaint thread on OpenAI's forum: almost all were negative, citing wrong answers, vague image limits, lost credits, outages and AI-only customer support. That's a self-selected, angry sample. But the forum post that says GPT-6 feels "even less intelligent than the models we had back in 2023" shows that even fans aren't universally won over.

Who it's for

OpenAI suits more people than any other AI company right now, simply because it covers so much ground:

  • Everyday users who want a free, capable assistant for questions, writing, homework help and planning. The free plan is enough for a lot of people.
  • Professionals who'll use the Work agent, Codex and Intelligent UI. Plus at $20 is the sweet spot.
  • Developers and businesses building on the API. Sol's price drop makes high-quality models far cheaper to ship.

Who should skip it? Anyone who needs reliable human customer support, anyone allergic to shifting usage limits, and anyone who wants to make video, since Sora is gone. If you rely on Agent Builder, OpenAI is effectively telling you to rebuild elsewhere on its platform.

OpenAI alternatives worth considering

  • Anthropic Claude: the strongest rival for writing and coding, with Opus 5.5 at or near the top of independent rankings. See our Claude AI review.
  • Microsoft Copilot: the natural pick if your work lives in Word, Excel and Outlook. Our Microsoft AI review explains how it fits together.
  • Google AI Mode: better for fast, search-grounded answers than long projects. Read our AI Mode review.

Verdict

OpenAI in October 2026 is still the default choice for a reason. ChatGPT is free to start, GPT-6 Sol and Luna are much cheaper to run than what came before, Intelligent UI is a genuinely new idea, and Codex keeps getting more capable. For developers, the API offers more models at more price points than anyone else.

But it's not a clean sweep. Independent tests show cost gains alongside some quality regressions, the plan lineup has become a maze of Pro tiers, Agent Builder is being shut down barely a year after launch, and the customer support complaints are loud and consistent. My advice: start free, move to Plus only when you hit the limits, and keep an eye on Claude, because the gap at the top is small. To get more from whichever assistant you pick, try our guide to writing better AI prompts.

FAQ

What is open AI?

OpenAI is the AI company that makes ChatGPT, the GPT-6 models (Astra, Sol and Luna), the Codex coding agent and the OpenAI API. It began as a nonprofit research lab in 2015 and now runs one of the world's most used consumer apps, with more than 1.2 billion weekly ChatGPT users.

Who owns open AI?

OpenAI's for-profit company is a public benefit corporation controlled by the nonprofit OpenAI Foundation. Microsoft held roughly 27% after the October 2025 restructuring, and investors in later rounds own shares too. Microsoft didn't buy OpenAI, and the company isn't publicly traded; Sam Altman has ruled out an IPO in 2026.

What does open AI do?

OpenAI researches and builds AI models, then sells access to them. Consumers use them through ChatGPT for chat, writing, images, coding and agent tasks. Developers and businesses pay per token through the OpenAI API to build the same models into their own apps, agents and services.

Is open AI free?

Yes, ChatGPT has a free plan with GPT-6 Luna, Intelligent UI, unlimited text chats within abuse limits and a few images a day. Paid plans start at $8 a month for Go and $20 for Plus. The OpenAI API isn't free: you pay per million tokens, from $0.10 input for Luna.

What video reviewers say

“In this video, I’ll be testing OpenAI’s new GPT-6.1 Sol model across all eight KingBench 3 tasks in Codex.”

— AICodeKing ·

“I Tested NEW GPT-6-Sol and GPT-6-Luna on 144 Coding Prompts”

— AI Coding Daily ·

Videos load from YouTube only when you press play.

Ratings around the web

Ratings as of 27 Oct 2026. Scores change over time; follow the links for the latest.

What users are saying

We read 24 recent user comments: the 20 most recent Trustpilot reviews we could access (late August to early September 2026) and the original post plus three replies in a GPT-6 complaint thread on OpenAI's forum (October 2026). This is a small, self-selected sample that skews far more negative than ChatGPT's 4.8-star app-store rating suggests.

Common praise

  • Answers almost anything you ask
  • Explains hard topics clearly, such as maths homework
  • Still seen by some critics as the best overall LLM

Common complaints

  • Wrong or inconsistent answers that need correcting
  • Vague image and usage limits, with credits lost to errors
  • Outages and AI-only customer support that rarely resolves billing issues

Based on: Trustpilot reviews of ChatGPT, OpenAI Developer Community: GPT-6 complaint thread

What we like

  • Free ChatGPT plan with GPT-6 Luna, Intelligent UI and unlimited text chats within abuse limits
  • GPT-6 Sol costs a fifth of Astra in the API, with much lower cost per task than GPT-5.6 Sol
  • Broadest product range: ChatGPT, Codex, Work agent, Agents API and a full developer platform
  • Intelligent UI turns answers into charts, forms and interactive tools
  • About 1.05M-token context window across the GPT-6 family

What could be better

  • Independent tests show some quality regressions on professional tasks alongside the cost gains
  • Agent Builder and Evals shut down on November 30, 2026, barely a year after launch
  • Confusing plan lineup with three Pro tiers and shifting usage allowances
  • Frequent user complaints about wrong answers, image limits, billing and AI-only support

Sources

  1. GPT-6 and Intelligent UI for everyone (OpenAI)
  2. OpenAI adopts new business model, keeps Microsoft (The Daily Ittefaq)
  3. OpenAI reportedly seeks $30 billion at $1.4 trillion valuation (ProPakistani)
  4. OpenAI Plans To Nearly Double Its Workforce (Let's Data Science)
  5. OpenAI DevDay 2026 Recap for Developers (InfoQ)
  6. GPT-6 Astra pricing: every API tier and ChatGPT plan in 2026 (eesel)
  7. OpenAI GPT-6 Sol and Luna API prices (TechEconomy)
  8. ChatGPT Go vs Plus vs Pro (Oct 2026) (The AI Career Lab)
  9. Introducing AgentKit (OpenAI)
  10. GPT-6.1 Sol, GPT-6 Sol and Luna: Pricing, Benchmarks, Tests (ComputingForGeeks)
  11. GPT-6 Is Horrible: Worse Intelligence, Memory and UX (OpenAI Developer Community)
  12. ChatGPT reviews (Trustpilot)
  13. ChatGPT Android app versions (Tech Dev Notes)
  14. GPT-6.1 Sol (Fully Tested) + Dots & All DevDay Launches Explained (AICodeKing, YouTube)
  15. I Tested NEW GPT-6-Sol and GPT-6-Luna on 144 Coding Prompts (AI Coding Daily, YouTube)

Keep reading