🔍 Read the full analysis: The 5X Is A Subsidy, Not A Price: SemiAnalysis On AI Subscriptions on ThorstenMeyerAI.com
Get movie-night favorites 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 usage limits across major AI subscriptions by model and token type, then estimated their value at providers’ API list prices. Its analysis puts Claude’s mid-tier plans at roughly 5.4 to 5.6 times the API-equivalent value of comparable ChatGPT plans, but says recent price and limit changes are eroding that gap and heavy usage can make subscriptions costly to serve.
The analysis measures how much each provider’s usage meter moves for different token types, including cached and uncached input, cache writes and output. It then prices the estimated usage at each provider’s first-party API list rates. For an agentic coding workload, SemiAnalysis estimates that a $20 Claude Pro plan allows about $1,178 in API-priced Opus 5.5 usage, while a $20 ChatGPT Plus plan allows about $211 in GPT-6.1 Sol usage. At $100 and $200 tiers, it reports a similar ratio.
Those dollar totals are model- and workload-dependent. The tested agentic workload is overwhelmingly cached input, and GPT-6.1 Sol costs less per token than Opus 5.5. SemiAnalysis says the gap remains substantial when comparing raw token counts, but its headline figures are estimates based on metered limits and published API prices, not cash paid out to subscribers. At the frontier tier, its comparison is closer: it estimates a $200 plan’s Astra allowance at about $2,897 of API use, while Fable 5.1 would consume roughly half of a Claude plan’s limit at about $2,485.
The report says OpenAI recently halved the token allowances on its $200 plan, with lower limits applying immediately to new purchases and existing subscribers keeping their former limits until October 29. OpenAI also introduced a $500 tier, which SemiAnalysis estimates offers only about 21% more Astra than the former $200 plan. Anthropic cut API prices too: the report says Fable 5.1’s cache reads cost 75% less than Fable 5’s, without a corresponding increase in limits. Opus 5.5’s input, output and cache-read prices also fell, while reported allowances rose by different amounts across Pro and Max tiers.
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 Inference Costs
SemiAnalysis estimates that subscriptions make up about 10% of Anthropic revenue but can use more than 40% of its inference compute. On those rough estimates, subscriptions lower blended revenue per megawatt by about $36 million. The report says subscriptions are a larger share of OpenAI revenue, making the balance between plan prices, usage limits and compute costs relevant to both providers’ business models.
The analysis also models margins at different usage levels. Assuming 92% API gross margins and a subscriber who uses the full allowance, it estimates gross margins of about −369% for Opus 5.5 and about 1% for Fable 5.1. At 20% average utilization, its estimates rise to about 6% and 80%, respectively. These are scenario calculations, not reported company results. They show why the headline value of a plan can be costly if subscribers use large amounts of premium-model capacity, while lower usage or cheaper models can change the economics sharply.
For customers, the practical implication is that a plan’s advertised price alone does not tell them how much work it will support. Model-specific limits, usage windows, token mix and changes to API prices all affect the comparison. SemiAnalysis says OpenAI’s Pro plans do not impose a five-hour usage window, which may help users who want to spend more of their monthly allowance in bursts, though it does not erase the report’s estimated API-value gap.
As an affiliate, we earn on qualifying purchases.
Recent Model and Plan Changes
SemiAnalysis’s comparison reflects plan terms after recent changes at both providers. It says OpenAI’s GPT-6.1 Sol launch did not bring a higher usage limit, and that the $200 plan’s API-equivalent value fell by roughly 30% as the model’s cached-input price declined. The report describes OpenAI’s current Pro 100, Pro 200 and Pro 500 tiers as returning similar tokens per dollar, whereas the prior ladder offered larger per-dollar allowances at higher tiers.
Anthropic’s adjustments were mixed, according to the report. Fable 5.1 launched without higher token limits despite lower cache-read prices; Opus 5.5’s allowances rose about 20% on Max and 50% on Pro, which SemiAnalysis says did not fully offset its price reductions. In both cases, lower API prices can reduce the API-equivalent value of a fixed subscription allowance. That comparison does not mean subscribers receive less service in every practical use: outcomes depend on which model and token types they use.
As an affiliate, we earn on qualifying purchases.
Limits and Usage Remain Variable
The supplied source material does not include the full methodology, underlying meter readings or provider responses, so the estimates cannot be independently checked here. API-equivalent value is a comparison against list prices; it is not a measure of every subscriber’s actual consumption, provider cost or realized margin. SemiAnalysis’s margin figures also rely on stated assumptions about utilization and API gross margins.
It is unclear how long the current allowances will remain in place, whether usage limits will change again, or how the new $500 tier’s Ultrafast mode will perform in practice. SemiAnalysis says it is still testing the 300-token-per-second mode. The report’s figures also may not apply to workloads with a different balance of cached input, fresh input and output.
As an affiliate, we earn on qualifying purchases.
Further Metering Will Test the Estimates
SemiAnalysis says testing of OpenAI’s Ultrafast mode is ongoing. A fuller assessment would also depend on future provider updates to subscription allowances, model prices and usage rules, as well as disclosed methodology that allows readers to compare the meter readings with their own workloads. Until then, customers weighing plans can use the report as a model-specific estimate and check current plan limits before purchasing.
As an affiliate, we earn on qualifying purchases.
Key Questions
What did SemiAnalysis compare?
It measured usage limits across major AI subscriptions by token type, then estimated the API list-price value of the usage those limits permit for selected workloads.
How large is the reported Claude-to-ChatGPT value gap?
For its agentic coding workload on mid-tier models, SemiAnalysis estimates Claude plans provide about 5.4 to 5.6 times the API-equivalent value of comparable ChatGPT plans. The result depends on the models, token mix and prices used in the comparison.
What changed for OpenAI’s $200 plan?
SemiAnalysis says OpenAI halved its token allowances. New purchases get the lower limits immediately; existing $200 subscribers keep their former limits until October 29.
Does API-equivalent value show what a subscriber will use?
No. It prices a plan’s estimated maximum allowance at API list rates. Actual usage varies by subscriber, model and workload, and the estimate is not a cash payout or a direct measure of provider costs.
What remains to be tested?
SemiAnalysis says it is still testing OpenAI’s 300-token-per-second Ultrafast mode. The source material does not establish how long current limits will last or whether providers will adjust them again.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
