A flat-rate seat is a promotion, and the ChatGPT Pro plan cut shows promotions get repriced. Price your team’s AI at what the work would cost at metered rates before the next change lands.
A $200 ChatGPT plan now buys about half the tokens per model tier it used to, according to SemiAnalysis, and the number on the card hasn’t moved. This ChatGPT Pro plan cut is a plain warning for any firm budgeting on flat seats. People who bought before the cut keep the old limits until October 29. New buyers get the lower ones immediately. A seat price tells you what you pay, not what you get, and the vendor can change the second number any week.
Key points
- What did the ChatGPT Pro plan cut change? Per SemiAnalysis , OpenAI halved the tokens on the $200 plan for each model tier.
- Is Claude really five times better value? On one comparison, yes, and the multiple moves with the comparison.
- Who pays for the subsidy? You do, eventually. SemiAnalysis says subscriptions can take over 40% of Anthropic’s inference compute while producing about 10% of its revenue.
- What should a 50-person firm do before the next repricing? Start small. First, count the people on flat plans and price their real usage at metered rates.
What did the ChatGPT Pro plan cut change?
Per SemiAnalysis, OpenAI halved the tokens on the $200 plan for each model tier. It also cut cached-input prices on GPT-6.1 Sol, which pushed that tier’s API-equivalent value down by more than 50%. The new $500 plan gives only about 21% more of the top Astra model than the old $200 plan.
Metered rates are where the real number sits, as GPT-5.5 pricing shows once you count tokens per finished task.
Its headline feature is a faster-output option. The $100, $200 and $500 plans now give the same tokens per dollar on every model.
The size of the cut rests on SemiAnalysis, which tracks usage meters, and on The Register’s coverage. We found no OpenAI announcement. The Register says OpenAI didn’t respond to its request for comment.
Is Claude really five times better value?
On one comparison, yes, and the multiple moves with the comparison. SemiAnalysis converted each flat plan’s usage into metered API prices and found Anthropic’s plans gave about 5x the API-equivalent value on mid-tier models, Opus 5.5 against GPT-6.1 Sol. It cites about 4x elsewhere in the piece.
The Register’s write-up says SemiAnalysis puts a $20 Claude Pro seat at about 2.9 billion tokens a month against about 1 billion for ChatGPT.
Then the counterweight. Artificial Analysis, which tracks cost per completed task, puts Sol at $0.72 per task against $5.98 for Opus 5.5 at API rates, according to the same write-up. A bigger allowance of an expensive model isn’t automatically cheaper work. We wouldn’t switch vendors on the 5x headline, and the comparison is useful mainly for what it says about the flat plans themselves.
Who pays for the subsidy?
You do, eventually. SemiAnalysis says subscriptions can take over 40% of Anthropic’s inference compute while producing about 10% of its revenue. If that holds, flat plans are a customer-acquisition cost, and the labs will adjust them as their economics demand. OpenAI just did.
The Register puts the compute share at 42%. Dhruv Amin, co-founder of agent business Skydive, told The Register that “the labs typically subsidize their prosumer plans with subscription pricing massively.”
The sharpest detail is in SemiAnalysis’ method notes. It found a tiny A/B test on one account showing roughly 20% lower limits, and says that confirms providers can change limits silently. A cut can reach you before anyone announces it.
Seats at $20 are riding that subsidy. A company building on metered API calls isn’t, and the bills are real. The Register cites a July McKinsey report saying 93% of enterprises exceeded their AI budgets. Our pricing pieces on Opus 5.5’s effort setting and Microsoft’s metered Copilot show the same pressure from the metered side. And Big Tech’s off-balance-sheet AI spending is the reason to expect prices to bend toward cost.
What should a 50-person firm do before the next repricing?
Start small. First, count the people on flat plans and price their real usage at metered rates. Say ten people sit on $20 seats, which is $200 a month. If that same usage would cost $600 at metered rates, you hold a $400 monthly subsidy you don’t control.
The $600 is a made-up figure to show the method, not a benchmark.
Second, list the workflows you’d keep if the plan halved tomorrow. Third, put a name against every renewal date and note the cancellation period. Cost per successful task is the metric that survives a repricing, and our GPT-6 Astra pricing look and our piece on AI costs for business cover the metered end.
Does this matter if your team rarely hits the limits?
Probably not this month. For light users a $20 seat may never bite, and this fight is mostly over power-user plans. The risk grows with agents and coding tools, which burn through tokens quickly, so a team that drifts into them meets the new limits without warning.
If your usage sits far below any cap, skip the exercise and keep the seats.
SemiAnalysis’ full tracking data sits behind a paywall, and its method, converting usage meters to API-equivalent dollars, is reasonable, though other methods would give different totals. Plan details change often, so check the vendor’s pricing page before you buy. If OpenAI posted a note showing the cut was smaller than reported, we’d revise how hard we lean on the subsidy argument.
Put the renewal dates in a calendar this week. The October 29 date is the first one.
If your team runs agents rather than chat seats, capping what each agent can spend on tokens is the next cost to plan.
Frequently asked questions
Did OpenAI cut the ChatGPT Pro plan?
According to SemiAnalysis, yes. It says OpenAI halved the tokens per model tier on the $200 plan and added a $500 plan. Existing $200 subscribers keep the old limits until October 29.
Is Claude cheaper than ChatGPT for a business?
It depends on what you measure. SemiAnalysis says Anthropic’s plans give about 5x the API-equivalent value on mid-tier models, while Artificial Analysis finds GPT-6.1 Sol far cheaper per completed task at API rates.
How should a small business budget for AI subscriptions?
Price the work at metered rates as well as seat rates, and plan for the vendor to change limits. Note each renewal date and cancellation period.
The figures on plan limits come from SemiAnalysis and are attributed, not verified by us. Written and fact-checked on October 8, 2026. Sources are linked on the claims they support.
