🔍 Read the full analysis: SemiAnalysis Breaks Down The 5X Behind AI Subscriptions on ThorstenMeyerAI.com
Get office and shipping supplies delivered free — and shop member deals
- Fast, free delivery on millions of items
- Access to Prime Big Deal Days deals on October 6–7
- Prime Video, Amazon Music and more included
TL;DR
SemiAnalysis compared major AI subscriptions by measuring token allowances and pricing equivalent usage at API list rates. On a coding-agent workload, it estimates Claude plans provide about 5.4–5.6 times the API-equivalent value of comparable ChatGPT plans, while recent limit and price changes have shifted the economics. The report also estimates that subscriptions use a disproportionate share of Anthropic’s inference compute, but actual subscriber usage and future plan limits remain uncertain.
SemiAnalysis has measured usage limits across major AI subscriptions and estimates that Claude plans provide about 5.4 to 5.6 times the API-equivalent value of similarly priced ChatGPT plans on a coding-agent workload. The comparison matters because it links advertised subscription limits to the compute costs behind them, as both OpenAI and Anthropic have recently adjusted model prices and plan allowances.
The report compares GPT-6.1 Sol with Claude Opus 5.5, pricing each plan’s full monthly allowance at the providers’ API list rates. For plans costing $20 per month, it estimates $211 in API-equivalent usage for ChatGPT Plus and $1,178 for Claude Pro, a ratio of about 5.6 to 1. At $100, the estimates are $1,055 for ChatGPT Pro 100 and $5,725 for Claude Max 5x. At $200, they are $2,084 for ChatGPT Pro 200 and $11,726 for Claude Max 20x.
Those estimates are tied to a workload dominated by cached input: SemiAnalysis describes it as roughly 96.6% cached input, 2.6% cache writes, 0.4% fresh input and 0.3% output. The report says GPT-6.1 Sol costs less per token than Opus 5.5, which affects the dollar comparison, but says the gap also remains large when measured in raw tokens. These figures represent the value of a fully used allowance at list prices; they do not show what a typical subscriber consumes.
The comparison reflects changes to both services. SemiAnalysis says OpenAI roughly halved the token allowances on its $200 plan, with existing subscribers retaining previous limits until October 29 and new purchases receiving lower limits immediately. The company also introduced a $500 plan. According to the report, that tier offers about 21% more Astra usage than the old $200 plan, while its selling point includes an Ultrafast mode advertised at 300 tokens per second. SemiAnalysis says it is still testing that mode.
The 5x is a subsidy, not a price
SemiAnalysis metered the meters — every major AI subscription, token type by token type, converted to API list value. On the mid-tier models both labs call the daily driver, a Claude plan returns ~5–6× the API value of the matching ChatGPT plan. Real — and the least durable number in the report.
…and the plan is fully exhausted. One pool for every model.
…and the plan is only half used — Fable is capped at 50% of the limit, leaving the rest for Opus/Sonnet. That’s where the mid-tier gap compounds.
- $200 plan halved — Sol-class value down >50% (6.1 Sol cache price cut compounds it)
- Old limits kept until 29 October; new buyers cut immediately
- New $500 tier: only +21% Astra vs the old $200 — real draw is 300 TPS Ultrafast
- Ladder flattened: Pro 100/200/500 now identical per dollar; multipliers removed from pricing page
- In OpenAI’s favour: no 5-hour window on Pro plans — easier to use the full allowance
- Flat per-dollar value across all tiers, before and after
- New premium models placed at lower relative limits (Fable capped at 50%)
- Opus allowances raised ~20% (Max) / ~50% (Pro) with the 5.5 price cut — not enough to fully offset it
- Repeatedly walked back planned cuts earlier this year under pressure from OpenAI’s generosity
- Twelve months ago, OpenAI was the generous option. Positions swap.
Gross margin per plan, assuming 92% API gross margins. The subsidy lives almost entirely in Opus and Sonnet usage — Anthropic would already be near software-like subscription margins if everyone used only Fable. Subscriptions matter even more for OpenAI, where they’re a larger share of revenue.
Three identical subscriptions; one had ~20% lower limits. The provider (unnamed) confirmed an “extremely tiny” A/B test on limit balancing. Two lessons: limits can change silently, per account, at any time — and you won’t know without instrumentation. The usage bar is a percentage, not a contract.
If you’re choosing a plan this month for agentic coding on a mid-tier model, the report settles it: a Claude plan returns ~5–6× the API value of the matching ChatGPT plan. But a plan returning 58× its fee on a model served at a steeply negative margin for heavy users is a marketing budget with a usage meter. Value moves silently, gets A/B tested per account, and twelve months ago ran the other way. Use the subsidy while it exists — it’s genuinely large. Don’t build a cost model on it. Price workloads at API rates, keep a router between you and any one vendor, and benchmark open weights on your own hardware for steady volume. A deal you can’t verify isn’t a price. It’s weather.
Subscription Value Meets Compute Costs
The reported value gap is relevant to customers choosing plans, but SemiAnalysis argues the underlying compute economics may matter more. Its rough estimates put subscriptions at about 10% of Anthropic revenue while accounting for more than 40% of its inference compute. The report estimates this mix lowers blended revenue per megawatt by about $36 million. It says subscriptions represent a larger share of OpenAI revenue, though it does not provide the same specific revenue share in the supplied comparison.
These figures suggest that generous limits can act as a costly way to attract and retain users, particularly when subscribers use the most expensive models heavily. SemiAnalysis estimates that a fully utilized Opus 5.5 subscription could produce a gross margin of about −369%, assuming API gross margins of 92%. Its estimate for Fable 5.1 is about 1%. At 20% average utilization, the report estimates margins of about 6% for Opus and 80% for Fable. These are scenario calculations based on the report’s assumptions, not reported financial results from either company.
For customers, the practical value depends on which models they use, how much they use them and whether their work fits within any time-based limits. OpenAI’s Pro plans have no five-hour usage window, according to SemiAnalysis, which could benefit people who need to use a large share of their monthly allowance in a short period. The report judges that advantage insufficient to erase the estimated value gap, but says it can narrow the difference for bursty workloads.
As an affiliate, we earn on qualifying purchases.
How Recent Price Cuts Shifted Limits
API list prices and subscription allowances do not necessarily move together. SemiAnalysis says Anthropic cut Fable 5.1 cache-read prices by 75% compared with Fable 5, without increasing Fable’s token limits. It says Opus 5.5 prices for input and output fell 20%, while cache reads fell 60%, compared with Opus 5. Anthropic raised allowances by about 20% on Max and 50% on Pro, according to the report, but those increases did not fully offset the price reductions in its API-equivalent calculation.
SemiAnalysis says OpenAI made no allowance change when GPT-6.1 Sol launched, while a reduction in Sol’s cached-input price led the report’s estimated API-equivalent value of the $200 plan to fall by about 30%. Separately, the recent plan revision cut token allowances. The report says the change flattened OpenAI’s former tier progression: Pro 100, Pro 200 and Pro 500 now return identical tokens per dollar in its comparison. It also says OpenAI removed “5x more usage” and “20x more usage” multipliers from its pricing page.
At the frontier model level, SemiAnalysis describes limits as broadly similar. It estimates a $200 OpenAI plan would be exhausted after about $2,897 of GPT-6 Astra usage at list prices. A comparable Claude plan would be half-used after about $2,485 of Claude Fable 5.1 usage, because Fable can consume only half of the plan’s limit. The remaining allowance is available for Opus or Sonnet, where the report says the mid-tier comparison shifts in Claude’s favor.
Actual Usage and Future Limits
The API-equivalent values assume a subscriber uses the full monthly allowance and that usage is priced at list rates. The report’s scenarios therefore do not establish average customer consumption, actual subscription profitability or the precise compute cost of each account. The supplied material also does not identify the sample size, testing period or method used to measure every provider’s usage bar, limiting how independently readers can assess the comparison.
It remains unclear whether OpenAI or Anthropic will change allowances again as model prices and compute costs shift. The comparison also does not establish how broadly the measured coding-agent workload represents other uses, such as general chat or research. SemiAnalysis’s estimate that OpenAI escaped significant backlash after its plan change is an interpretation; the report points to next-day DevDay announcements and a delay before existing subscribers’ limits change as possible factors.
Limits Change on October 29
The next dated change in the report is October 29, when existing subscribers to the $200 ChatGPT Pro plan are due to lose their grandfathered limits. New purchases already receive the reduced allowances, according to SemiAnalysis. The report says it is continuing to test OpenAI’s 300-token-per-second Ultrafast mode, so its practical performance and value remain to be assessed.
Further changes to model prices or subscription limits could alter the API-equivalent comparisons. For now, the report offers a snapshot built around specified models, plan terms and a cached-input-heavy coding workload. Readers weighing a subscription can use its figures as estimates for that workload, while checking current plan limits and their own likely usage before drawing conclusions about value.
Key Questions
What did SemiAnalysis compare?
It measured usage allowances across several AI subscriptions and estimated what the full allowance would cost at each provider’s API list prices. Its main comparison uses GPT-6.1 Sol and Claude Opus 5.5 on a coding-agent workload.
How large was the estimated Claude advantage?
For the three paired tiers in the report, Claude’s API-equivalent value was estimated at about 5.4 to 5.6 times ChatGPT’s. The estimate assumes full use of the plan allowance and applies to the workload tested.
Did OpenAI change ChatGPT Pro limits?
SemiAnalysis says OpenAI roughly halved allowances on its $200 plan. New purchases receive the lower limits immediately; existing subscribers are due to retain their prior limits until October 29.
Do the estimates show what subscribers usually use?
No. They calculate the API list-price value of a fully used allowance. The report’s figures do not establish typical subscriber usage or actual profit and loss per account.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
