I spend a good chunk of every week inside ChatGPT, comparing it against the other assistants I recommend on this site, so I keep close tabs on what OpenAI changes and how often it changes it. And 2026 has been a genuinely confusing year for anyone trying to figure out which plan makes sense.

OpenAI split Pro into two price points, pushed a new flagship model family into the lineup, put ads on the free plans, and just extended those ads into 31 European markets days before I sat down to write this. None of the “ChatGPT pricing” articles floating around right now have caught up. This one has.

The Full 2026 ChatGPT Lineup, Side by Side

Here’s the whole board. I built this from OpenAI’s current pricing pages and Help Center articles, not from a comparison post someone wrote back in the spring.

A quick note on the model names, because OpenAI hasn’t made this easy. GPT-5.6 landed in three flavors on July 9, 2026: Sol (the flagship), Terra (a mid-weight model tucked into ChatGPT Work and Codex), and Luna (the fast, cheap default). Luna took over as the Free and Go default on August 6, replacing GPT-5.5 Instant. If you see an older article still talking about GPT-5.3 or GPT-5.4 as the current models, it’s out of date. So is anything that still lists Sora as dead weight on Pro — more on that below.

How I Judged Each Tier

I don’t run this comparison based on a single afternoon of clicking around. I pulled every claim here from OpenAI’s own pricing and Help Center pages, cross-checked the numbers against multiple independent trackers, and weighed each plan against the workflows I actually see readers of this site running: content production, client work, coding side projects, and small-team operations.

Where a feature is genuinely in flux — and several are, this year — I’ll flag that instead of pretending the picture is settled. OpenAI itself has said pricing will keep evolving, and the still-unreleased Business Premium seat tier is proof of that in real time.

Free: What $0 Actually Buys You Now

Free stopped being simple the moment ads showed up. They landed in the US in February 2026, and as of August 24 they’re running across 31 European markets too — buyable through OpenAI’s self-serve Ads Manager, which tells you this isn’t a test anymore, it’s a business line.

Here’s the trade I didn’t expect: OpenAI removed the old message cap. Free users are moving to unlimited text chats on GPT-5.6 Luna as the rollout completes, following the model switch on August 6. That’s a real improvement over the old “10 messages then a silent downgrade” system that used to define the free tier.

What you still don’t get: the flagship Sol model in standard chat, Deep Research, Sora, Agent Mode, or a context window bigger than roughly 16,000 tokens. That last number matters more than people realize — it’s the difference between ChatGPT remembering your whole conversation and it quietly forgetting the first half.

My take: Free is fine for someone who opens ChatGPT twice a week to draft an email or settle an argument. The second it becomes part of your actual job, the missing context window and locked-out features start costing you more time than the $20 would.

ChatGPT Go: A Plan I Wouldn’t Recommend to Most People

Go went global in January 2026 at $8 a month, after starting as a roughly $4–5 India-only experiment. It’s now sold in about 98 countries. On paper it sits neatly between Free and Plus. In practice, I think it’s the weakest value on this entire list.

Here’s my problem with it: Go runs the same GPT-5.6 Luna model as Free, gets the same unlimited-text upgrade Free just received, and still shows the same ads. What you’re actually paying $8 for is higher file and image caps, a longer memory window, and access to build (not just browse) Custom GPTs. You still don’t get Deep Research, Sora, Agent Mode, or the flagship Sol model.

So the real comparison isn’t Go versus Plus. It’s Go versus Free, and once both tiers show ads and both get unlimited text, the case for paying $8 gets thin fast. If you’re outside the US where ad coverage still varies, Go holds up a little better. In the US, I’d tell you to save the $96 a year and either stay Free or jump straight to Plus.

ChatGPT Plus: The Plan I’d Actually Tell You to Buy

Plus is still $20 a month, monthly billing only, and it remains the single most popular paid AI subscription for a reason. It’s the cheapest tier where the flagship GPT-5.6 Sol model shows up in standard chat, selectable at medium and higher reasoning effort since July 9, 2026. GPT-5.5 Instant sticks around as the fast default for quick questions, with Sol and Thinking available when a task actually needs the extra reasoning.

The context window jumps to roughly 320 pages of text per conversation, which is the single biggest quality-of-life upgrade over Free and Go if you work with long documents, contracts, or multi-turn research threads.

And here’s a correction worth making, because a lot of pricing roundups still get this wrong: Sora didn’t disappear from Plus. OpenAI shut down the standalone Sora app, website, and API in March 2026 after the short-form video product failed to gain lasting traction and a reported Disney partnership fell through. But the underlying video model stayed inside ChatGPT. Plus subscribers still get limited Sora generation (up to 5 seconds at 720p) baked into the chat interface itself.

What else you’re unlocking at $20: 10 Deep Research runs a month for cited, multi-source reports, the full Codex agent for reviewing and debugging code on GitHub, Agent Mode for autonomous multi-step tasks, Canvas for collaborative writing and code editing, Advanced Voice Mode, and a completely ad-free interface.

If you’re weighing Plus against building your own custom assistants, I broke down that exact decision when I compared Claude Projects against ChatGPT’s Custom GPTs, and it’s worth a read before you commit to one ecosystem.

Pro $100 vs Pro $200: OpenAI’s Confusing Middle Ground

This is the part of the lineup that still trips people up, and honestly, I think OpenAI designed it that way on purpose. On April 9, 2026, OpenAI split its old $200 Pro tier by launching a $100 entry point, aimed squarely at Anthropic’s Claude Max, which has held the same $100 price for over a year.

Here’s the part that surprises most people: both Pro tiers give you the exact same model access. Pro $100 and Pro $200 both include GPT-5.6 Sol Pro, the higher-accuracy variant reserved for Pro and Enterprise, plus o1 Pro mode for harder reasoning tasks. The only real difference is usage headroom. Pro $100 gives you 5x Plus’s limits. Pro $200 gives you 20x, along with a full 1-million-token context window, unlimited Deep Research, and unlimited Sora and image generation.

Worth flagging: Pro $100 launched with a promotional 10x Codex boost over Plus, but that promo expired on May 31, 2026, so don’t go in expecting it — it’s long gone by the time you’re reading this.

My honest opinion: unless you’re a developer who lives in Codex sessions or a researcher who burns through Deep Research runs before the week is out, neither Pro tier is worth it. Pro $100 makes sense the moment you’re hitting Plus’s usage ceiling on a normal workday. Pro $200 only makes sense if you’re running several demanding tasks in parallel, not just working faster — working wider. For everyone else, that $80–180 a month is better spent elsewhere. If you’re on the fence, I already walked through the full math in my is ChatGPT Pro worth $200 a month breakdown.

Business and Enterprise: When It Stops Being a Personal Decision

Business (formerly called Team, before an August 2025 rename) costs $25 a seat billed monthly, or $20 a seat billed annually, with a 2-seat minimum. It includes everything in Plus, unlimited access to GPT-5.6 Sol, shared workspaces, admin controls with SAML/SSO, and integrations with Slack, Google Drive, SharePoint, and GitHub.

The number that actually matters here isn’t the feature list, it’s the training default. Free, Go, Plus, and both Pro tiers all use your conversations to train future models unless you manually flip the setting off in Data Controls. Business and Enterprise skip that step entirely — training is off by default, no toggling required. For any team touching client data, that alone justifies the jump.

One thing to watch: OpenAI announced a Business Premium seat on August 10, 2026, priced at $125 a month ($100 annually), offering 5x the usage of a standard seat and removing the 5-hour usage cap that cuts off long agent runs. It’s real, but it’s currently waitlist-only. As of my last check, it still wasn’t listed on the public pricing page, so treat it as announced, not available.

Enterprise sits above all of it at custom pricing, requires roughly 150 or more users, and adds SCIM provisioning, extended 128K+ context, data residency options, and 24/7 SLA-backed support. If you’re at the size where you’re negotiating this directly with OpenAI’s sales team, you already know it’s a different conversation than anything on this page.

Data Privacy by Tier: Who Trains on Your Prompts

If privacy is the deciding factor for you and not the budget, that’s a genuinely underrated reason to go Business even as a two-person operation. I’d also recommend reading how this compares against Claude’s approach before you lock in — I covered the accuracy side of that comparison in Claude vs ChatGPT hallucination rates.

My Verdict: Which Plan Should You Actually Buy

If you use ChatGPT a few times a week for casual questions: stay on Free. The unlimited-text change genuinely closes the gap that used to make this tier frustrating. Just budget for the ads.

If ChatGPT touches your actual work — writing, research, code, marketing: buy Plus. It’s not close. The Sol model, the 320-page context window, Deep Research, and the ad-free interface make the $20 pay for itself within a week for most professionals.

If you regularly slam into Plus’s limits: try Pro $100 before you jump to $200. Most people who think they need the top tier actually need the middle one.

If you’re running a team of two or more, especially with client data: Business is the obvious pick. Training is off by default, and the per-seat cost lands close to Plus once you’re billing annually.

And a real disclosure: I run several AI subscriptions side by side for this site’s testing, and Plus is the one I’d tell a friend to buy first, no hesitation. Everything above it is a “grow into it” purchase, not a starting point.

Before you commit either way, it’s worth seeing how a rival flagship stacks up in a real workflow — I put GPT-5.6 through its paces against Claude in Claude Fable 5 vs ChatGPT 5.5, and the gap between them isn’t as wide as the marketing suggests.

FAQ

Is ChatGPT Plus worth $20 a month in 2026?

For most working professionals, yes. Plus unlocks the flagship GPT-5.6 Sol model, a roughly 320-page context window, Deep Research, Codex, Agent Mode, and an ad-free interface. Skip it if you use ChatGPT fewer than a couple of times a week.

Does ChatGPT Free have ads now?

Yes. Ads appear under responses on Free and Go, live in the US since February 2026 and expanded to 31 European markets on August 24, 2026. Plus, Pro, Business, and Enterprise remain ad-free.

What’s the actual difference between Pro $100 and Pro $200?

Both include the same GPT-5.6 Sol Pro model and o1 Pro mode. Pro $100 gives you 5x Plus’s usage limits; Pro $200 gives you 20x, plus unlimited Deep Research, Sora, and image generation.

Is Sora gone from ChatGPT?

The standalone Sora app and API shut down in March 2026. The video generation model itself stayed inside ChatGPT — Plus gets limited generation, Pro $200 gets unlimited.

Is ChatGPT Go worth it over Free?

Rarely, in my view. Both tiers now run the same model and both show ads. Go’s extra file, image, and memory allowances aren’t enough to justify $8 a month unless you’re outside the US where ad rollout is patchier.

Does ChatGPT offer student pricing?

Not as an ongoing individual discount. OpenAI runs an institutional ChatGPT Edu plan for universities instead. If student pricing across AI tools matters to you, I compared the landscape in my Perplexity Pro student discount guide.

Conclusion

OpenAI has turned ChatGPT’s pricing page into a moving target, and I don’t expect that to stop. But strip away the model-name confusion and the Business Premium seat that isn’t even shipping yet, and the actual decision is simple: Free now holds up better than it used to, Go isn’t worth your money, Plus is the plan that earns its keep for almost everyone doing real work, and both Pro tiers are specialist purchases, not defaults.

Buy for what you need this quarter. Given how often OpenAI has reshuffled this lineup in 2026 alone, revisit the decision in six months rather than assuming today’s plan will look the same by then. If you’re building out a broader AI toolkit for content or marketing work, my guide to using ChatGPT for small business marketing and my roundup of AI SEO tools are both good next stops, and if agent-style tools are on your radar, it’s worth checking whether Claude Cowork is safe to use before you commit to one ecosystem over another.