ChatGPT Free vs Plus vs Pro: What You Actually Get at Each Tier (Tested)

ChatGPT Free costs nothing and now gives you unlimited everyday text chats. Plus costs $20 a month and is the first tier that feels like the full product: stronger reasoning models, expanded deep research, projects, scheduled tasks, custom GPTs, and no ads. Pro is for people whose work depends on ChatGPT all day — it comes in three tiers, Pro 100 at $100, Pro 200 at $200, and Pro 500 at $500 a month, and the extra money buys much larger usage allowances, the top Pro reasoning models, and longer coding and research sessions. Most people should start free, upgrade to Plus the week they keep hitting a limit, and only consider Pro when Plus is genuinely slowing their work down.

ChatGPT is OpenAI’s conversational AI assistant, and its subscription plans are simply different sizes of the same tool: Free is the trial-sized version, Plus is the daily-driver version, and Pro is the heavy-work version. The plans differ less in what the assistant can do in principle and more in how much of it you can use, how capable a model answers you, and which tools are switched on. This guide tests those differences tier by tier, using OpenAI’s own pricing page and help centre as the source of truth.

Checked 6 November 2026. Prices and limits in this article are current as of that date, based on OpenAI’s pricing page and help centre, read in full. OpenAI changes plans, models, and limits often — the Pro tier changed three times in the autumn of 2026 alone — so check the price at checkout before you pay. Where OpenAI does not publish a number, this article says so instead of guessing one; figures marked “reported” come from independent trackers that cite OpenAI’s pages, not from OpenAI directly.

1. The Short Answer: Which Tier for Whom

<img src="

” alt=”A person at a desk choosing between three blank plan cards laid out in a row” />

If you use ChatGPT a few times a week for quick questions, short drafts, and the odd explanation, Free is enough — and it is genuinely free, with no card required and unlimited everyday text chats. The moment you start using it daily, upload documents, generate images, or run research, you will feel the edges of the free tier within a week, and that is the moment Plus earns its $20. Plus is the right tier for most individuals: students, writers, freelancers, professionals, and anyone who treats ChatGPT as a daily work tool rather than an occasional curiosity.

Pro is a different kind of purchase. At $200 a month for the middle tier, it costs ten times what Plus costs, and it only makes sense when your income or your output depends on heavy AI use: researchers running deep research all day, developers living in Codex, analysts working with large documents, or teams of one producing at industrial volume. If you have to ask whether you need Pro, you almost certainly do not — yet. The honest test is simple: if Plus never makes you wait, Pro will not make you faster. If Plus makes you wait every day, Pro is worth a one-month trial.

There is also a fourth personal plan most comparisons skip. ChatGPT Go sits between Free and Plus, aimed at people who want more room than Free without paying for Plus. It gives you more uploads, more images, more voice chats, and longer memory, but it keeps the free tier’s model and its ads. It is a reasonable choice for light-but-regular users, though in the United States its price is only a few dollars a month and the exact figure varies by region, so confirm it at checkout. For a full walkthrough of what the product itself can do on any tier, our complete guide to using ChatGPT covers the features from first prompt to power-user workflows.

2. The Three Tiers at a Glance

The table below is the whole comparison in one place. Every row comes from OpenAI’s pricing page or help centre, checked 6 November 2026. Where OpenAI does not publish a number, the table says “limited” or “not published” rather than inventing one.

*Unlimited everyday text chat is subject to abuse guardrails and a reasonable-use policy on every plan. The asterisks matter: “unlimited” covers plain text conversation, not the tools around it. Uploads, images, voice, deep research, and the heavier models all have their own limits, and those limits are where the tiers actually separate. Read the table as a map, not a verdict — the sections below test each row in practice.

Two things in the table surprise most people. First, the free tier is far more capable than it was a year ago: unlimited text chat, web search, image generation, voice, and some deep research, all for nothing. Second, the gap between Plus and Pro is mostly about volume and the very top models, not about everyday features — Plus already includes projects, custom GPTs, scheduled tasks, and deep research. That is why the upgrade decision is really a usage decision, and why the rest of this guide is organized around what each tier feels like to use.

3. Free Tier, Tested: What You Get and What You Lose

Free is a real product, not a demo. You get unlimited everyday text chats with GPT-5.6 Luna, OpenAI’s fastest and lowest-cost model, plus web search, image creation, voice chats, file uploads, data analysis, deep research, and Codex — each in limited amounts. You can try almost everything ChatGPT does before spending anything, which is exactly what the tier is designed for. For occasional use — a few questions a day, a short email draft, a recipe, a homework explanation — Free may never ask you for money at all.

What you lose is headroom, and the losses arrive in a predictable order. The first wall most people hit is file uploads: three files per day, full stop. Attach three documents to your first prompt and you are done uploading until the allowance resets. The second wall is image creation, which is capped at a number OpenAI does not publish; you will simply be told you have reached the limit. The third is voice: limited voice chats, and no advanced voice features. Deep research is available but limited — independent trackers report around five research runs a month, though OpenAI does not publish the figure. Memory and context are limited too, so ChatGPT remembers less about you across sessions than it does on a paid plan.

The model is the subtler loss. Free runs on GPT-5.6 Luna, and its Think mode also runs on Luna. Luna is fast and capable for everyday work, but it is the lightest model in the family: it reasons less deeply than the GPT-6 Sol models that Plus and Pro users get, and it has no access to the Medium, High, Extra High, or Pro thinking levels. For simple questions you will not notice. For a hard reasoning problem — a tricky piece of logic, a long document to untangle, a genuinely difficult piece of analysis — the difference shows up as shorter, shallower answers and more frequent mistakes. Free users also see ads, which paid plans do not, and ChatGPT Work — the mode that creates documents, spreadsheets, and presentations — is limited to the desktop app on Free.

Free is also the tier where small frustrations compound. Projects hold only five files each. Library storage is 500 MB. Audio file uploads are not available on Free at all. And when you hit any tool limit, the only immediate fix is to wait — or upgrade, which resets the free limits at once. None of this makes Free a bad tier. It makes it a tier with a clear job: try the product, use it lightly, and pay when your use outgrows it. If you find yourself rationing uploads or waiting out image limits more than once a week, you have outgrown it.

4. Plus Tier, Tested: What $20 a Month Actually Buys

Plus is the tier where ChatGPT stops feeling like a free sample and starts feeling like a tool you own. For $20 a month you get everything in Free and Go, plus the upgrades that matter most to daily users: the GPT-6 reasoning models, expanded limits on nearly every tool, projects, scheduled tasks, custom GPTs, expanded Codex, expanded ChatGPT Work on desktop, web, and mobile, early access to new features, and no ads. It is, by a wide margin, the best-value tier OpenAI sells — and the tier this guide recommends to most readers.

The single biggest upgrade is the model. Plus users get GPT-6, powered by GPT-6 Sol, with Medium and High thinking levels you can choose by hand. Thinking levels are how much effort the model spends reasoning before it answers: Medium for ordinary hard questions, High for the ones that need real care. The difference from Free’s Luna is not subtle on difficult work. Answers get longer, more structured, and more reliable; multi-step problems that Luna fumbles, Sol works through. You do not get Extra High or the Pro thinking level — those are reserved for Pro — but for most professional and study work, Medium and High are the levels you will actually use.

The second upgrade is headroom. File uploads jump from 3 a day to up to 80 every three hours. Projects hold 25 files instead of 5. Library storage rises from 500 MB to 20 GB. Image creation becomes more capable, including image generation with thinking — a slower, more careful mode for images that need to be right. Deep research expands substantially; independent trackers report around 25 runs a month, though OpenAI does not publish the number. Voice expands too, and Plus adds video and screen sharing in voice conversations. Memory and context expand, so ChatGPT remembers more about you and holds more of a long conversation in view.

The third upgrade is the workflow features. Projects let you organize chats, files, and instructions under one objective — a job search, a course, a book, a business plan — so you stop re-explaining the same background. Scheduled tasks let ChatGPT do things proactively on a schedule, like a morning briefing or a recurring check. Custom GPTs let you build specialized assistants tuned for one job, with their own instructions and knowledge files. These are the features that turn ChatGPT from a chat window into a system, and they are the reason Plus feels so different from Free even before you count the model upgrade. Add the removal of ads and early access to new features, and $20 a month is easy to justify for anyone who opens ChatGPT most days. The hard question is not whether Plus is worth it — it is whether you use ChatGPT enough to notice the difference, and the next section helps you answer that.

5. Pro Tier, Tested: What $200 (and $100, and $500) Actually Unlocks

Pro is not one plan any more. Since the autumn of 2026, OpenAI sells three monthly Pro tiers: Pro 100 at $100, Pro 200 at $200, and Pro 500 at $500. Pro 200 is the tier most people mean when they say “ChatGPT Pro,” and it is the tier this section tests — but the existence of the other two matters, because Pro 100 is now the sensible entry point for heavy users and Pro 500 is the only tier with Ultrafast mode. All three are billed monthly; annual billing is not available for Pro. If you were following the Pro story earlier in 2026, note that OpenAI canceled the Pro 200 usage reduction it announced in September, and Pro 200 is available for new subscriptions again.

What does the money buy? Everything in Plus, plus the top of the range: Pro reasoning powered by GPT-6 Astra, longer Codex and ChatGPT Work sessions, Dot — OpenAI’s always-on agent — unlimited and faster image creation, maximum deep research, maximum memory and context, expanded projects, tasks, and custom GPTs, research previews of new features, and monthly API credits for building your own apps and agents. The Pro tiers differ from each other mainly in how much usage they include: Pro 200 includes more than Pro 100, and Pro 500 offers the highest included usage of the three, plus Ultrafast mode for GPT-6 Astra and GPT-6.1 Sol, which spends your allowance faster in exchange for lower latency. OpenAI has also doubled the Work and Codex allowances on Pro 100 and Pro 500, and doubled the lower Pro 200 allowance to match the higher one.

The honest assessment is that Pro’s value is narrow and deep. For a researcher running deep research all day, a developer living in Codex, or an analyst working through hundreds of pages, the difference between Plus and Pro is the difference between working and waiting: no five-hour usage window on Work and Codex, much larger allowances, and the Pro reasoning models for the hardest questions. For everyone else, Pro is $100–$500 a month spent on headroom they will never use. The Pro models themselves are capped even on Pro — OpenAI is explicit that a Pro subscription does not include unlimited use of the GPT-6 Pro model — so Pro does not remove limits, it moves them much further away. That is worth $200 a month only if you regularly reach the Plus limits. OpenAI’s own history makes the point: Pro limits have changed repeatedly, and our earlier report on OpenAI doubling Pro limits and dropping a planned cut shows how quickly the deal can move in either direction.

The practical way to think about Pro is as a business expense, not a subscription. If ChatGPT is how you earn your living — research, code, analysis, content at volume — then $200 a month is a small fraction of a professional’s costs, and the time saved by never hitting a limit pays for itself in days. If ChatGPT is how you study, write, or plan, Plus already gives you the models and the tools; Pro gives you more of them, and more is not better when you were not using what you had. Try Pro 100 first if you are curious: it is half the price of Pro 200, includes the same Pro models, and will tell you within a month whether the extra headroom matters to you.

6. Model Access Compared: GPT-6, Thinking Levels, and the Pro Models

The model is the engine, and the tiers give you different engines. As of November 2026 the current family is GPT-6, built on the GPT-5.6 line: Luna is the fastest and lowest-cost model, Terra balances capability, speed, and cost for everyday work, and Sol is the reasoning model that powers the GPT-6 experience for paid users. Free and Go users get GPT-5.6 Luna as their default, and their Think mode also runs on Luna. They do not get access to GPT-5.6 Sol at all. Plus users get GPT-6, powered by GPT-6 Sol, with Medium and High thinking levels. Pro, Business, and Enterprise users get Medium, High, Extra High, and Pro thinking levels. That table, from OpenAI’s own help centre, is the single most important fact in this comparison — because the model you get shapes every answer you receive.

Thinking levels deserve a plain explanation, because OpenAI’s naming invites confusion. A thinking level is how much reasoning effort the model spends before answering. Instant is the fast, light answer. Think is a slower, more careful answer. Medium and High are progressively deeper reasoning for harder questions. Extra High and Pro are the deepest reasoning OpenAI offers in chat, reserved for Pro and business plans. Higher thinking does not make the model smarter in the abstract; it gives the model more room to work through a hard problem step by step, check its own reasoning, and catch its own mistakes. On easy questions, higher thinking is wasted effort and slower answers. On hard questions, it is the difference between a plausible answer and a correct one. If you want the broader story of how these models relate to each other, our OpenAI explainer, from GPT-3 to GPT-5 traces the company’s model history, and our guide to choosing an AI model helps you match a model to a task.

The Pro models are the top of the range: GPT-6 Pro in chat, powered by GPT-6 Astra, and GPT-5.6 Sol Pro. They are the models Pro subscribers reach for on the hardest questions, and they are capped even on Pro — OpenAI states plainly that a Pro subscription does not include unlimited use of the GPT-6 Pro model. When you hit the weekly GPT-6 Pro allowance on Pro 200, you fall back to GPT-5.6 Thinking at Medium. OpenAI does not publish the exact weekly counts in the help article read for this guide, so this article does not state them; the practical point is that even $200 a month does not buy unlimited access to the very best model. That is worth knowing before you pay, because the marketing around Pro can leave the opposite impression.

One more thing the model table quietly tells you: the free tier is not running an old model. Luna is a current model, and for everyday text chat it is genuinely good. The gap between Luna and Sol is real but narrow for simple work and wide for hard work, which is exactly the shape of a well-designed free tier. If your questions are mostly simple — explanations, short drafts, summaries, translations — Luna will serve you well and you will never feel short-changed. If your questions are mostly hard — research, analysis, code, long documents — the Sol models on Plus will feel like a different product. The model is the reason Plus is worth $20, and it is also the reason Free is worth trying first.

7. Message and Usage Limits Compared

The biggest change in ChatGPT’s limits happened in August 2026, and it is good news for everyone: everyday text chat is now unlimited on every plan, including Free, subject to abuse guardrails and a reasonable-use policy. There is no longer a message count on plain text conversation. The old world — ten messages every five hours on Free, then a fallback to a smaller model — is gone. If you only ever type questions and read answers, you will never hit a limit on any tier. That single fact reshapes the whole comparison: the tiers no longer differ in how much you can chat, only in how much of everything else you can do.

Everything else still has a limit, and the limits live in three places. First, the thinking levels: choosing Medium or High reasoning uses the Sol model, and that usage is capped by plan — Plus gets Medium and High with no published count, Pro gets all levels including Pro, also with no published count in the help article read for this guide. When you hit a thinking limit, ChatGPT tells you and shows when the allowance resets. Second, the Pro models in chat: GPT-6 Pro and Sol Pro have weekly and daily allowances on Pro tiers, and OpenAI does not publish the counts. Third, ChatGPT Work and Codex share an allowance measured in rolling five-hour windows on Plus and Business Standard — OpenAI publishes estimated ranges that vary by model and by how much work each message causes, from a handful of messages for the heaviest models to thousands for the lightest. Pro tiers have no five-hour limit on Work and Codex, which is one of the clearest practical differences between Plus and Pro.

There are also limits that behave like speed bumps rather than walls. File uploads, image generation, voice, and data analysis each have their own caps, covered in the sections below. Long conversations can hit a context limit — the amount of text the model can hold in view at once — and when that happens ChatGPT may summarize or forget earlier parts of the conversation. OpenAI’s pricing page notes that the context window is shared between your input, system instructions, memories, and the model’s own processing, and that the space available for your input is an approximation that changes with the features in use. No token numbers per plan are published on the pricing page, so this article does not state any.

The practical advice is simple and it comes from OpenAI itself: when you hit a limit, ChatGPT shows you when it resets. Do not try to memorize the limits — they change, and they vary with demand, model, and task. Instead, notice which limit you hit. If you hit the same one every week, that limit is telling you which tier you need. A free user who hits the upload cap daily needs Plus. A Plus user who hits the Work and Codex window daily needs Pro. And a useful piece of context: paid accounts have had per-card and per-device caps too, as our report on ChatGPT capping paid accounts per card and device covered — limits are a moving target, and the checked date on this article is there for exactly that reason.

8. File Uploads, Memory, Projects, and Custom GPTs

Files are where the free tier’s limits bite first and hardest. Free users get three file uploads per day — documents, spreadsheets, and images all count — and audio uploads are not available on Free at all. Paid plans can upload up to 80 files every three hours, a rolling rate rather than a daily cap. Every plan shares the same per-file ceiling of 512 MB, and only ten files can go in at the same time. If you work with documents regularly — reports, contracts, research papers, spreadsheets — the free tier’s three-a-day cap will feel absurd within a week, and it is the single clearest reason to pay for Plus. These figures come from OpenAI’s File Uploads FAQ as cited by multiple independent trackers; the FAQ itself was not opened in full for this guide, so treat the counts as well-corroborated rather than first-hand.

Memory is the second quiet differentiator. Free has limited memory and context, Go has longer memory, Plus has expanded memory and context, and Pro has the maximum. In practice, memory is what lets ChatGPT remember your preferences, your projects, and the facts you have told it across sessions, so you stop re-explaining yourself. On Free, that memory is thin: fewer remembered details, less context carried between conversations. On Plus and Pro, it is thick enough to feel like working with an assistant who knows you. You control it in Settings under Personalization, and you can ask ChatGPT what it remembers about you, tell it to remember something, or tell it to forget something. The difference between tiers is not whether memory exists — it is how much of it there is.

Projects, scheduled tasks, and custom GPTs are the workflow features, and they draw the sharpest line between Free and Plus: the pricing page lists them under Plus, not Free. Projects organize chats, files, and instructions under a shared objective, with project-only memory available to keep a project’s context sealed off. A project holds five files on Free, 25 on Go and Plus, and 40 on Pro. Scheduled tasks let ChatGPT act proactively — a morning briefing, a recurring check, a reminder. Custom GPTs are specialized assistants you build with tailored instructions and knowledge files, shareable or publishable to the GPT Store; note that memory is not currently available inside custom GPTs, per OpenAI’s own documentation. Together these three features are what turn ChatGPT from a chat window into a system for repeatable work, and they are only fully available from Plus upward.

The verdict on this section is straightforward. If you never upload files, never build a workflow, and never want ChatGPT to remember anything about you, Free is fine. If you upload documents, keep long-running projects, or want an assistant that remembers your preferences, Plus is not a luxury — it is the tier where those features exist at all. Pro adds more room: more files per project, more storage (100 GB of library storage against Plus’s 20 GB), and expanded projects and tasks. But the jump from Free to Plus is a jump from “not available” to “available,” while the jump from Plus to Pro is a jump from “available” to “more available.” That is why Plus is the value tier and Pro is the volume tier.

9. Voice, Images, and Video per Tier

Voice is one of ChatGPT’s most underrated features, and the tiers treat it very differently. Free gets limited voice chats. Go gets more voice chats. Plus gets expanded voice, including video and screen sharing inside voice conversations. Pro gets expanded voice as well. OpenAI does not publish minute limits for any tier, so this article does not state any — but the shape of the difference is clear: on Free, voice is a taster; on Plus, it is a usable daily feature; on Pro, it is the same feature with more room. Voice Mode is a two-way real-time conversation — you talk, ChatGPT replies out loud, and you can interrupt — while dictation simply turns your speech into text you can edit before sending. Audio clips from voice chats are stored with the transcription in your chat history for as long as the chat remains there.

Image generation follows the same pattern. Free gets limited image creation. Go gets more. Plus gets more complex and accurate image creation, including image generation with thinking — a slower mode that reasons about the image before making it. Pro gets unlimited and faster image creation. Again, OpenAI publishes no per-day counts, so there is no number to quote and this article will not invent one. What matters in practice is the experience: free users hit the image limit without warning and have to wait; Plus users rarely hit it; Pro users effectively never do. Image input — uploading a photo, a screenshot, a chart, or a whiteboard and asking questions about it — is available on all tiers within the upload limits, so the free tier is not locked out of visual work entirely. It is locked out of doing a lot of it.

Video is the murkiest part of the comparison, and honesty requires saying so. OpenAI’s current pricing page does not list Sora, its video-generation tool, as a feature of any ChatGPT plan. Independent trackers report that the standalone Sora app closed in April 2026, but that is not verified from an OpenAI page in this session, so this guide does not present Sora as a reason to choose any tier. What Plus does include, according to the same trackers, is video and screen sharing inside voice conversations — showing ChatGPT what your camera sees, or sharing your screen while you talk. That is a genuinely useful feature for troubleshooting, learning, and working through visual problems, and it starts at Plus. If video generation is the feature you care about most, check OpenAI’s current pages before paying for any tier on that basis — the situation has changed more than once.

The practical takeaway: voice and images are the features that make ChatGPT feel effortless, and effortlessness is exactly what the free tier rations. If you talk to ChatGPT while cooking, driving, or walking, or if you generate images for work, the free tier’s limits will annoy you daily and Plus will feel like a release. If you mostly type, the voice and image limits will rarely matter, and your upgrade decision should turn on the model and the file limits instead.

10. Deep Research and Agents per Tier

Deep research is ChatGPT’s most powerful feature for knowledge work, and the tiers ration it aggressively. Deep research reads and synthesizes content across many online sources and produces a cited, structured report — a literature review, a strategy document, a market scan — in minutes rather than hours. Free gets limited deep research. Go gets limited deep research. Plus gets expanded deep research. Pro gets maximum deep research. OpenAI does not publish the monthly counts, so this article gives the figures reported by independent trackers with an explicit caveat: around five runs a month on Free, around 25 on Plus, and far more on Pro. Treat those as reported, not verified — OpenAI may change them without notice, and the in-product counter is the only authoritative source for your account.

The quality of deep research does not change much between tiers; the quantity does. A free user gets a taste: enough to run one serious research task a month and discover the feature exists. A Plus user gets enough for a weekly research habit — a report a week, a literature scan before a big decision, a competitive scan before a launch. A Pro user gets enough to make deep research a daily tool. The lesson is the same as everywhere else in this comparison: the tiers differ in volume, not in kind. If you need deep research once a month, Free might stretch far enough. If you need it once a week, Plus is the tier. If you need it every day, you are a Pro customer whether you have admitted it yet or not. For the craft of turning AI research into content people actually cite, our GEO explainer on getting cited by ChatGPT and AI Overviews is the natural companion to this section.

Agents are the newer frontier, and the tier picture is still settling. ChatGPT’s agent mode can browse the web, use tools, and complete multi-step tasks on your behalf. Codex, the coding agent, is available in limited form on Free and Go, expanded on Plus, and with longer sessions and no five-hour limit on Pro. ChatGPT Work — the mode that creates documents, spreadsheets, and presentations and works in a browser — is limited to the desktop app on Free and expanded across desktop, web, and mobile on Plus. Pro adds Dot, OpenAI’s always-on agent, which the pricing page lists as a Pro-only feature. The usage pool for Codex and Work is shared, and OpenAI has been merging the two — our report on ChatGPT and Codex sharing one usage pool covers that change. As with deep research, the practical rule is volume: agents are available on every tier, but the amount of agent work you can do grows sharply with the price.

One caution applies to all of this. Deep research and agents are the features most likely to be oversold by their own marketing, because a cited report feels authoritative even when it is not. Cited does not mean correct. A deep research report is a strong first draft, not a verified fact — check the citations for the claims that matter, look for missing perspectives, and ask follow-ups that probe the weakest section. That advice holds on every tier, but it matters most on Pro, where the volume of research output can outrun the time you have to verify it.

11. Cost-per-Use Reasoning: Doing the Maths Before You Pay

The right way to choose a tier is not to compare feature lists — it is to compare your usage against the price, and the arithmetic is kinder than most people expect. Start with Plus at $20 a month, or $240 a year. If you open ChatGPT twenty days a month, that is $1 per working day. If you open it every day, it is about 65 cents a day. If it saves you thirty minutes of work a week — a draft written, a document summarized, a problem solved — it has paid for itself many times over at any plausible value of your time. The question is never really “is Plus worth $20?” It is “do I use ChatGPT enough for $20 to matter?” And for anyone who uses it most days, the answer is yes before the maths starts.

Pro needs the same treatment with bigger numbers. Pro 200 at $200 a month is $2,400 a year — ten times Plus. At that price, the question is not whether Pro is better than Plus; it is whether the extra $180 a month buys you $180 of extra output. For a professional whose billing rate is $50 an hour, Pro pays for itself if it saves about three and a half hours a month over Plus — the time you would otherwise spend waiting on limits, splitting work across sessions, or falling back to weaker models. For a researcher who runs deep research daily, that saving is real and easily measured. For a student or a casual user, it is imaginary: the extra headroom sits unused, and $180 a month is a lot to pay for headroom.

Three worked examples make the decision concrete. First, a freelance writer who uses ChatGPT daily for drafting, editing, and research: Free would ration her uploads and images within days, Plus gives her the Sol models, the file headroom, and the projects to organize her clients — Plus is the obvious tier, and Pro would add nothing she uses. Second, a graduate student writing a thesis: deep research a few times a month, long documents, occasional image generation — Plus covers it, and the money saved over Pro buys books. Third, a data analyst at a small firm who runs deep research and Codex sessions all day, every day: Plus would throttle him by mid-morning, Pro 200 removes the five-hour window and multiplies the allowances, and the $200 is a rounding error against his salary. The same tool, three different answers — because the right tier is a function of your usage, not of the product.

Two money-saving notes before you pay. First, try before you buy at each step: Free is free, and a single month of Plus will tell you more about your usage than any comparison article — including this one. Subscribe for one month, use ChatGPT the way you actually work, and notice which limits you hit. Second, remember that Pro has no annual billing and no discount for paying ahead, while Business plans do offer annual pricing — if you are buying for a team, the business plans may be cheaper per seat than individual Pro subscriptions, and section 12 covers them. And always check the price at checkout: prices vary by country and currency, and the figures in this article are US list prices checked 6 November 2026.

12. Team, Business, and Enterprise in Brief

So far this guide has covered personal plans. If you are buying for a group, the maths changes, and OpenAI’s business plans deserve a brief look — brief, because most readers of a Free vs Plus vs Pro comparison are buying for themselves, and the business plans are a separate decision with their own pricing page.

ChatGPT Business is the self-serve plan for small teams, formerly called ChatGPT Team. A Standard seat costs $25 per user per month on a monthly plan, or $20 per user per month billed annually, with a minimum of two seats. A Premium seat costs $125 monthly or $100 billed annually, and gives five times the usage of a Standard seat with no five-hour limit. Business includes the Plus-level features — GPT-6 models, deep research, projects, custom GPTs, Codex — plus team administration, shared workspaces, SAML SSO and MFA, and a promise that OpenAI will not train on your workspace data. It does not include API usage, which is billed separately. A Standard seat gets Plus-level Work and Codex usage; a Premium seat gets five times that. These figures come from OpenAI’s Business pages as cited by multiple independent trackers.

ChatGPT Enterprise is the sales-led plan for large organizations, and OpenAI does not publish a price for it. You contact sales, and the price, seat types, and billing follow your agreement — with credit-based and token-based pricing both available. Enterprise adds what large companies need and small teams do not: SCIM provisioning, advanced admin controls, audit logs and usage monitoring, data residency controls, custom data-retention terms, Zero Data Retention and BAAs on contract, SLAs, 24/7 priority support, and invoice billing with purchase orders and net terms. Enterprise workspaces draw on a shared credit pool for advanced models and features. There is also ChatGPT Edu, a near-twin of Enterprise for schools, with similar administration and no public price.

The decision rule for teams is simple. If you are two to two hundred people who want ChatGPT with shared administration and data controls, start with Business — it is self-serve, the price is published, and a Standard seat costs the same as a Plus subscription. If you need invoicing, purchase orders, SCIM, data residency, or SLAs, you need Enterprise, and the conversation starts with OpenAI’s sales team rather than a checkout page. And if you are one person buying for yourself, ignore this section entirely: the personal plans are cheaper and simpler, and Business only makes sense when the team features matter.

13. Should You Upgrade, Downgrade, or Stay? (Decision Guide)

Stay on Free if you use ChatGPT a few times a week, mostly for short questions and quick drafts, and you have never hit a limit. Free is genuinely good now — unlimited text chat, a current model, and a taste of every tool — and there is no reason to pay for headroom you do not use. Stay on Free also if you are still deciding whether ChatGPT fits into your life at all: use it for a month, notice what frustrates you, and let the frustrations choose your tier for you. If nothing frustrates you, you have your answer.

Upgrade to Plus when you hit the same free limit twice in a week — the three-uploads-a-day cap, the image limit, the voice limit, the shallow answers on hard questions. Upgrade also when you start a project that needs ChatGPT to remember context across sessions, organize files, or run scheduled tasks, because those features start at Plus. And upgrade when you use ChatGPT daily: at $20 a month, daily use makes Plus one of the cheapest professional tools you own. The upgrade is reversible — you can cancel through the platform where you purchased, and access normally continues to the end of the billing period — so a one-month trial costs $20 and tells you everything.

Upgrade to Pro when Plus is slowing your work down every day: you hit the Work and Codex five-hour window, you run out of deep research, you need the Pro reasoning models for the hardest questions, or you need longer coding sessions. Start with Pro 100 at $100 — it includes the same Pro models as Pro 200 — and move up only if you outgrow it. Pro 200 is the middle tier and the right choice for most heavy users; Pro 500 is for people who need the highest usage and Ultrafast mode, and at $500 a month it is a professional tool for professional volumes. If you are weighing ChatGPT against a rival rather than against its own tiers, our complete Claude vs ChatGPT comparison tests the alternative directly.

Downgrade when the usage that justified the tier disappears. Finished the thesis? Drop from Plus to Free and see whether you miss it. Changed jobs and no longer live in Codex? Drop from Pro to Plus. The tiers are monthly, the switch takes minutes, and there is no penalty for moving down — your chats and history stay with your account. The most expensive mistake in this whole comparison is not choosing the wrong tier; it is paying for a tier you chose six months ago and stopped needing three months ago. Set a calendar reminder for the month after any upgrade: if you cannot name what the tier bought you, downgrade.

14. Frequently Asked Questions

Is ChatGPT Free really free? Yes. The Free plan costs $0, requires no card, and now includes unlimited everyday text chats with GPT-5.6 Luna, subject to abuse guardrails. Uploads, image creation, voice, deep research, memory, and Codex are limited on Free, and ads may appear. You can use Free indefinitely without paying.

Is ChatGPT Plus worth $20 a month? For anyone who uses ChatGPT most days, yes. Plus buys the GPT-6 reasoning models with Medium and High thinking, much larger limits on uploads, images, voice, and deep research, projects, scheduled tasks, custom GPTs, expanded Codex and ChatGPT Work, early access to new features, and no ads. For occasional users, Free is enough and Plus is not worth it.

What is the difference between ChatGPT Plus and Pro? Plus is $20 a month; Pro comes in three tiers at $100, $200, and $500 a month. Pro adds the Pro reasoning models (GPT-6 Pro, Extra High thinking), longer Codex and ChatGPT Work sessions with no five-hour limit, maximum deep research, memory, and context, unlimited and faster image creation, the Dot always-on agent, monthly API credits, and research previews. Plus already includes the core workflow features — projects, tasks, custom GPTs, deep research — so the gap is mostly volume and top models, not everyday features.

Which ChatGPT model do I get on each plan? Free and Go get GPT-5.6 Luna, with Think running on Luna. Plus gets GPT-6, powered by GPT-6 Sol, with Medium and High thinking levels. Pro, Business, and Enterprise get Medium, High, Extra High, and Pro thinking levels, plus access to the GPT-6 Pro model. OpenAI’s help centre publishes this table; free and Go users do not get access to GPT-5.6 Sol.

Is there a message limit on ChatGPT? Not for everyday text chat: since August 2026, text chat is unlimited on every plan, including Free, subject to abuse guardrails. Limits apply to thinking levels, the Pro models, file uploads, image generation, voice, deep research, and ChatGPT Work and Codex — not to plain conversation. When you hit a limit, ChatGPT shows when it resets.

How many files can I upload to ChatGPT? Free users are limited to 3 file uploads per day. Paid plans can upload up to 80 files every 3 hours. The per-file ceiling is 512 MB on every plan, and up to 10 files can be attached at once. Projects hold 5 files on Free, 25 on Go and Plus, and 40 on Pro. These figures come from OpenAI’s File Uploads FAQ as cited by independent trackers.

Does ChatGPT Free have deep research? Yes, in limited form. OpenAI does not publish the monthly count; independent trackers report around five runs a month on Free and around 25 on Plus. Pro gets the maximum allowance. The in-product counter is the authoritative source for your account.

Can I cancel ChatGPT Plus or Pro at any time? Yes. Cancel through the platform where you purchased — chatgpt.com, the Apple App Store, or Google Play — and you normally keep access until the end of the billing period. Uninstalling the app does not cancel the subscription. Pro is billed monthly only; annual billing is not available for Pro.

Is there a ChatGPT plan between Free and Plus? Yes — ChatGPT Go. It gives you more uploads, images, voice chats, and longer memory than Free, but keeps the free tier’s model (GPT-5.6 Luna) and its ads. The US price is reported as a few dollars a month and varies by region, so confirm it at checkout.

Do ChatGPT plans include Sora video generation? Not according to OpenAI’s current pricing page, which does not list Sora as a feature of any ChatGPT plan. Do not choose a tier on the basis of video generation without checking OpenAI’s current pages first.

Which plan should a student choose? Start with Free. If you hit the upload or deep research limits during term, Plus at $20 a month covers a thesis, a research project, or a heavy study load — and it is easy to cancel in the holidays. Pro is almost never the right choice for a student unless your research depends on daily deep research or Codex use.

Can I switch plans later? Yes. You can upgrade or downgrade between personal plans in Settings, under My Plan, and the change takes effect as confirmed at checkout. Business workspaces are separate: you can keep a personal Pro subscription alongside a Business workspace if you keep the workspaces separate.

Free is enough for most people, Plus is worth $20 for anyone who uses ChatGPT daily, and Pro is worth $200 only when your work depends on heavy, daily AI use. Start free, pay attention to which limit you hit first, and let that limit — not a feature list — choose your tier. Check the price at checkout before you pay, because OpenAI changes plans often, and revisit the decision every few months: the tier you need today may not be the tier you need next term.