ChatGPT Pro Plan Change: Is the $200 Tier Still Worth It?

The ChatGPT Pro plan at $200 a month is open to new subscribers again, but the usage behind it now nets out at half the old value in API-dollar terms. OpenAI argues cheaper models close the gap. That's true if you mostly use GPT-6 Sol or Luna, and less true if you lean on Astra. Teams running automated work should compare the plan against pay-as-you-go API pricing before they buy.

Key Takeaways

• OpenAI paused new $200 Pro sign-ups on September 10 because of demand for GPT-6 Astra, and reopened them this week.
• The new allowance nets out at half the old plan's value in API-dollar terms, OpenAI's Codex lead said. The price stays at $200.
• OpenAI says the five-hour limit won't return, and that falling API prices will make each dollar of allowance do more work.
• That offset is real for GPT-6 Sol and Luna, which cost about half as much as the models they replaced, but it doesn't help much if most of your work runs on Astra.
• For automated jobs and team pipelines, the API is often the better fit. Track a month of real usage before you pick.

What Changed in the ChatGPT Pro Plan?

The price stayed the same. The amount of work it buys changed.

What OpenAI said: In a post on X, OpenAI's Codex lead, Thibault Sottiaux, known as Tibo, announced the reopening. He wrote that OpenAI is "changing how we calculate the usage," and added: "In effect, if you do the math, it will net out at half the dollar in API spend compared to the old Pro $200 plan."

He made four commitments alongside it:

• No five-hour limit. OpenAI won't bring it back, so you can use your weekly allowance whenever you want.
• More work over time. OpenAI says subscribers will get more done per dollar as models get more efficient and API prices fall.
• No inflated list prices. OpenAI says it won't raise API prices to make the plan look like a better deal.
• Extras that don't use the allowance. New subscription features that won't count against usage, which he didn't name.

The post doesn't say whether the new counting applies to people who already had the $200 plan, or only to new subscribers. If you're an existing subscriber, check your plan page and usage dashboard rather than assuming.

Why Did OpenAI Pause the $200 Tier?

Demand for its newest model outran its computing capacity.

Reported fact: OpenAI paused new Pro sign-ups on September 10, 2026, a week after launching GPT-6 Astra, TechCrunch reported. Sottiaux said, "Demand for Astra is really unprecedented." Existing Pro subscribers kept their plans, and other tiers stayed open.

So the reopening comes with a trade. New sign-ups are back, but each subscription carries half the old API-dollar value. That's a way to spread limited capacity across more people.

Heavy users had already been feeling the squeeze. In a Hacker News discussion of the announcement, some Pro users said they were using up their weekly allowance within a day or two of heavy Astra work. Those are individual reports, not measured data.

What Does the ChatGPT Pro Plan Lineup Look Like Now?

Four paid tiers, and a new top end.

Plan	Price per month	What we know about usage
Plus	$20	The base paid tier
Pro $100	$100	5 times Plus's Codex usage, per OpenAI's April launch
Pro $200	$200	Was 20 times Plus; now half the old value in API dollars
Pro 500 (new)	$500	The highest allowance, plus the Ultrafast speed tier

The Pro $100 tier launched in April 2026, with OpenAI saying it offered "5x more Codex usage than Plus," per its community announcement. OpenAI hasn't published a new multiplier for the $200 plan after the change, so treat any "10 times Plus" estimate you see as a guess.

Pro 500 arrived at OpenAI's DevDay on September 29, carrying the highest usage allowance of any plan and access to Ultrafast, CNBC reported.

Is the ChatGPT Pro Plan Still Worth $200?

It depends on which models you use. That's the part most coverage skips.

OpenAI's argument is that API prices keep falling, so half the API-dollar value can still buy as much work as before. For some models, that's true today:

Model	API price per million tokens (input / output)	Change
GPT-6 Luna	$0.10 / $0.50	About half of GPT-5.6 Luna
GPT-6 Sol	$2 / $10	Half of GPT-5.6 Sol
GPT-6.1 Sol	$2 / $10, cached input $0.10	Same as GPT-6 Sol, with cheaper caching
GPT-6 Astra	$10 / $50	Five times GPT-6.1 Sol

GPT-6.1 Sol and Astra prices come from VentureBeat's DevDay report. We covered the Sol and Luna price cuts in our GPT-6 Sol and Luna guide.

Here's the catch. The models that got cheaper are Sol and Luna. Astra, the model that drove the demand, costs five times as much as GPT-6.1 Sol. If most of your work runs on Astra, you now get half the API-dollar allowance, and there's no matching price cut to offset it.

Our view: for Sol-heavy users, OpenAI's argument roughly holds, because the models they use got about 50% cheaper. For Astra-heavy users, this is closer to a straight halving.

GPT-6.1 Sol may narrow that gap: OpenAI says it matches Astra on one coding test at a fifth of the price, per VentureBeat. That's OpenAI's own evaluation. Test it on your own work.

Who Is Hit Hardest by the ChatGPT Pro Plan Change?

Not everyone on the $200 plan will notice. The change bites hardest where people already pushed against the limits.

• Astra-heavy power users. They get half the API-dollar allowance, and Astra didn't get cheaper. Some were already running out of weekly usage in a day or two, by their own accounts.
• People running agents in parallel. Several Codex sessions at once burn through an allowance fast. Halving it shortens how long that setup lasts each week.
• New subscribers. The announcement clearly covers them. Whether it covers existing $200 subscribers isn't stated.

Others may barely feel it:

• Light and medium users who rarely hit a limit won't see much difference.
• API customers pay per token, so the plan change doesn't touch their bills.
• Sol and Luna users get cheaper models that stretch the smaller allowance further.

The honest read: OpenAI's framing that you "will still get more work done" than a month ago can be true and still feel like a cut, depending on your mix of models. Watch your own usage for a few weeks before you decide it's worse, or better.

Subscription or API: How Should a Team Decide?

By how you work, and by what you can measure.

A subscription suits one person working interactively. You get a predictable bill and don't have to think about tokens. The API suits automated work: agents, pipelines, batch jobs and anything that runs without a person watching.

• Choose a subscription if one person does most of the work in Codex or ChatGPT, the usage is steady, and a fixed monthly bill matters more than precise cost tracking.
• Choose the API if work runs in scripts, agents or pipelines, several people share it, or you need costs per project or per customer.
• Mix both if developers code interactively on subscriptions while production jobs run on the API.

The API also has two cost levers a subscription hides. Cached input on GPT-6.1 Sol costs $0.10 per million tokens, 95% below the standard input price. And routing lets simple jobs run on Luna at a twentieth of Sol's price. Our Sonnet 5.5 vs Luna comparison shows how much routing can change the bill.

How Do You Do the Math?

OpenAI hasn't published what the $200 allowance is worth in dollars. So you can't calculate the exact break-even from its announcement alone. You can still estimate it from your own usage.

• Log a month of real work. Count input tokens, cached tokens and output tokens per task, and note which model ran each one.
• Price that usage at API rates. Multiply each token type by the model's price per million.
• Compare with the plan. Well below $200 at API rates means the API is cheaper. Far above it means a plan may win, as long as you don't keep hitting its limits.
• Count the hidden costs. Waiting for a limit to reset costs developer time. So do retries.

Here's a simple illustration, not a measurement. Say a coding task on GPT-6.1 Sol uses 400,000 input tokens, of which 300,000 are cached, plus 40,000 output tokens. That costs about $0.03 for cached input, $0.20 for fresh input and $0.40 for output, so roughly $0.63 per task. At that rate, $200 buys about 317 tasks a month on the API.

Now run the same task on GPT-6 Astra, at $10 input, $1 cached input and $50 output per million tokens. It costs about $3.30, so $200 covers only about 60 tasks. That five-fold gap is why the model you use matters as much as the plan you pick.

If your team runs far fewer tasks than that, the API is likely cheaper. If you run far more on Astra, check the plan's limits closely. Our guide to token efficiency explains why tokens per task, not price per token, drives the bill.

What Else Changed at OpenAI DevDay?

A lot, and some of it affects the value of a plan.

• GPT-6.1 Sol. A new model at $2 input and $10 output per million tokens, rolling out in ChatGPT and Codex.
• Ultrafast. A premium speed tier that OpenAI says generates tokens up to eight times faster in Codex and up to six times faster in the API. It's priced at six times standard rates, per VentureBeat, and it's available now for Astra, with Sol "coming soon."
• Pro 500. The new $500 plan with the highest allowance and Ultrafast access.
• Dots. Always-on agents that run in the background on GPT-6 Astra, Digital Trends reported.
• Plugin extensions. Apps that developers can build to run inside ChatGPT and Codex.

OpenAI also said a day earlier that it had pulled plans to launch GPT-6.1 Astra because it did not meet its safety standards, CNBC reported. That fits the wider safety debate we covered in our look at AI agent incidents at the UN Security Council.

What Should Teams Do Now?

Treat this as a pricing change and review it like one.

• Existing Pro subscribers: check whether your allowance changed, and watch your usage dashboard for a few weeks.
• New subscribers: start on the tier you think you need, then move up only when limits actually block work.
• Teams on the API: set per-project budgets, turn on caching and route simple jobs to cheaper models.
• Everyone: keep a second model ready. Capacity limits, pauses and price changes can arrive with little notice, as this month showed.

Our guide to AI agent evaluation covers how to test a second model on real tasks, so a switch doesn't catch you unprepared.

How Van Data Team Helps Teams Control AI Spend

We help teams track what their AI work actually costs and pick the cheapest setup that does the job. That means usage logging per task, model routing, caching and budgets for agents and pipelines, and a clear view of when a subscription or the API makes more sense.

The ChatGPT Pro plan change is a reminder that AI pricing moves fast. If you want help working out your own numbers, our AI governance guide is a good place to start.