🔍 Read the full analysis: SemiAnalysis Examines The Subsidy Behind AI Subscription Pricing on ThorstenMeyerAI.com
Get monitors, keyboards and dev gear 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 the usage limits of major AI subscriptions by token type and converted them into estimated API list-price value. Its analysis found Claude’s mid-tier plans offer about 5.4 to 5.6 times the API-equivalent usage of comparable ChatGPT plans in the tested agentic workload, but the figures depend on model, usage and changing limits.
SemiAnalysis has published a comparison of usage allowances across major AI subscriptions, estimating how much the included tokens would cost at each provider’s API list prices. For a tested agentic workload, it put comparable Claude mid-tier plans at roughly 5.4 to 5.6 times the API-equivalent value of ChatGPT plans, while documenting recent limit reductions at OpenAI and price and allowance changes at Anthropic.
The analysis tested subscriptions from Anthropic and OpenAI, alongside offerings from Meta, SpaceXAI, Cursor, Cognition, Z.ai, MiniMax and Moonshot. SemiAnalysis measured how usage meters changed across token types, then applied first-party API list prices to estimate the dollar value of a plan’s stated monthly allowance. Its central comparison uses Claude Opus 5.5 and GPT-6.1 Sol on an agentic workload dominated by cached input: the report describes the workload as roughly 96.6% cached input, with smaller shares of fresh input, cache writes and output.
In that comparison, SemiAnalysis estimated API-equivalent value of $1,178 for Claude Pro at $20 per month, against $211 for ChatGPT Plus. At the $100 level, it estimated $5,725 for Claude Max 5x and $1,055 for ChatGPT Pro 100. At $200, the estimates were $11,726 for Claude Max 20x and $2,084 for ChatGPT Pro 200. These are estimates of the full monthly allowance priced at API rates, not cash refunds or guaranteed value for every subscriber.
The report says the gap remains large when usage is compared in raw tokens, rather than API dollars, because GPT-6.1 Sol is cheaper per token than Opus 5.5. It also finds a narrower comparison at the frontier tier: GPT-6 Astra and Claude Fable 5.1 have broadly similar stated limits, though Claude’s allowance is shared across models. SemiAnalysis says Fable can use only half of a Claude plan’s limit, leaving the remainder for Opus or Sonnet.
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.
The Cost of Subsidized Usage
The comparison matters because an attractive monthly allowance can represent a substantial inference-compute subsidy, particularly when subscribers use the most expensive models heavily. SemiAnalysis estimates subscriptions account for about 10% of Anthropic’s revenue but may consume more than 40% of its inference compute. It calculates that this mix lowers blended revenue per megawatt by roughly $36 million. The report says subscriptions represent a larger share of OpenAI’s revenue, though it does not provide a directly comparable figure in the supplied analysis.
SemiAnalysis’s margin estimates show how strongly usage patterns shape the economics. Assuming a subscriber uses the full allowance and that API business gross margins are 92%, it estimates maxing out Opus 5.5 would imply a gross margin of about negative 369% on the subscription; maxing out Fable 5.1 would yield about 1%. At an assumed 20% average utilization, its estimates rise to about 6% for Opus and 80% for Fable. These are modelled scenarios, not reported company-wide subscription margins.
For customers, the practical question is not only which plan looks more generous on paper. It is whether the chosen model, usage window and limits fit their work—and whether those limits will remain stable. A plan’s API-equivalent value can fall when a model’s API price drops, even if the subscription’s token allowance does not change.
As an affiliate, we earn on qualifying purchases.
Recent Changes to Plan Limits
SemiAnalysis’s comparison reflects a recent OpenAI change that the report says halved allowances on its $200 plan. Its tracking found token limits per model tier were roughly halved. The report says the API-equivalent value for Sol-class models fell by more than half because OpenAI also reduced GPT-6.1 Sol’s cached-input price. Existing $200 subscribers keep their old limits until October 29, while new purchases receive the lower limits immediately, according to the analysis.
OpenAI also introduced a $500 tier. SemiAnalysis estimates it provides about 21% more Astra usage than the former $200 plan, and less Sol-class API value. The report identifies 300 tokens per second in “Ultrafast” mode as the tier’s main distinguishing feature, but says it was still testing that mode. OpenAI also removed “5x more usage” and “20x more usage” multipliers from its pricing page. The report says the remaining Pro plans now provide similar tokens per dollar, though none has a five-hour usage window.
Anthropic has reduced API prices too. SemiAnalysis says Fable 5.1 cut cache-read prices by 75% compared with Fable 5, while Opus 5.5 reduced input and output prices by 20% and cache reads by 60% compared with Opus 5. It reports no corresponding increase in Fable’s subscription limits; Opus allowances rose about 20% on Max and 50% on Pro, still short of fully offsetting the price cuts in API-equivalent terms.
As an affiliate, we earn on qualifying purchases.
Limits of the Value Estimates
The reported ratios depend on one selected workload, model prices and plan limits. SemiAnalysis’s agentic test is heavily weighted toward cached input, so results for other tasks—with different proportions of fresh input, cache writes and output—may differ. API-equivalent value also does not establish how much an individual subscriber will actually use or save.
The analysis relies on the providers’ published list prices and measured usage meters. The supplied material does not provide the full test protocol, sample size, or independent verification of the providers’ internal compute costs. Its revenue and margin figures are estimates based on stated assumptions, not audited financial disclosures. It is also unclear whether all listed providers were tested on directly comparable plans or workloads.
Some plan details remain in motion. SemiAnalysis said it was still testing OpenAI’s Ultrafast mode, and the material does not establish how long current limits will last or whether providers will make further adjustments. The figures should be read as a snapshot, not a forecast or permanent ranking.
As an affiliate, we earn on qualifying purchases.
Future Plan Changes to Watch
The next near-term date identified in the report is October 29, when existing subscribers to OpenAI’s $200 plan are due to lose their grandfathered limits, according to SemiAnalysis. New customers already receive the reduced allowance. The report’s assessment of the $500 plan may also change as testing of its Ultrafast mode progresses.
For subscribers comparing plans, the useful next step is to check the current limits for their specific model and tier rather than relying on a single headline ratio. Future changes to API list prices can alter the calculated subscription value even if monthly fees stay the same. Any provider increase—or reduction—in included usage could shift the comparison again.
As an affiliate, we earn on qualifying purchases.
Key Questions
What did SemiAnalysis compare?
It measured subscription usage limits by token type and estimated the API list-price cost of using the full allowance. The report covered major offerings from OpenAI and Anthropic and several other AI providers.
What does the reported 5.4-to-5.6-times gap mean?
For the tested agentic workload using GPT-6.1 Sol and Claude Opus 5.5, SemiAnalysis estimated that Claude’s mid-tier plans provide about 5.4 to 5.6 times the API-equivalent value of comparable ChatGPT plans. It is not a claim that every user or task receives that same practical benefit.
Did OpenAI reduce its $200 plan limits?
According to SemiAnalysis, OpenAI roughly halved token allowances across model tiers on that plan. The report says existing subscribers keep old limits until October 29, while new purchases receive the reduced limits immediately.
Does a lower API price increase subscription value?
Not automatically. If a model’s API price falls but a subscription’s token allowance stays fixed, the allowance is worth less when calculated at the new API rate. SemiAnalysis says this occurred with some recent model and plan changes.
Are the subscription margin figures company disclosures?
No. They are SemiAnalysis estimates based on assumptions about API gross margins and subscriber utilization. The report’s scenarios should not be treated as audited or company-reported subscription margins.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
