🔍 Read the full analysis: Why The 5X Figure May Be A Subsidy In AI Subscriptions on ThorstenMeyerAI.com
Get the latest gadgets 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 estimates that, on a coding-agent workload, Claude’s mid-tier subscriptions provide about five to six times the API-equivalent value of comparable ChatGPT plans. The report also tracks recent changes to OpenAI and Anthropic plans and estimates that heavy use of some premium models can make subscriptions costly to serve. The estimates depend on workload, prices and usage limits, which providers can change.
SemiAnalysis has compared the token allowances in major AI subscriptions with the cost of buying equivalent usage at API list prices. For a coding-agent workload using mid-tier models, it estimates that Claude plans deliver about five to six times the API-equivalent value of similarly priced ChatGPT plans. The report says the gap comes with a cost: heavy use of some premium models can make subscriptions expensive to serve.
The comparison covers Claude, ChatGPT and plans from Meta, SpaceXAI, Cursor, Cognition, Z.ai, MiniMax and Moonshot. SemiAnalysis tracked how each plan’s usage meter moved across token types, then priced the measured allowance at first-party API rates. Its central comparison uses Claude Opus 5.5 and GPT-6.1 Sol on an agentic workload, where most input tokens are cached. The resulting dollar figures are estimates of API-equivalent usage, not cash paid out to subscribers.
At the $20 tier, SemiAnalysis estimates $211 of API-equivalent GPT-6.1 Sol usage for ChatGPT Plus and $1,178 of Opus 5.5 usage for Claude Pro, a ratio of about 5.6 to one. It reports similar ratios for the $100 and $200 tiers. The report says the difference remains large when measured in raw tokens, though the models’ different per-token prices affect the dollar comparison. At the frontier tier, the allowances are closer: the report estimates that a $200 plan provides about $2,897 of GPT-6 Astra usage, while Claude’s limit is half-used by about $2,485 of Claude Fable 5.1 usage. The other half remains available for Opus or Sonnet.
The comparison reflects a recent OpenAI plan change. SemiAnalysis says the company roughly halved token allowances across model tiers on its $200 plan. Because GPT-6.1 Sol’s cached-input price also fell, its estimated API-equivalent value dropped by more than half. Existing $200 subscribers retain their previous limits until October 29; new purchases receive the lower limits immediately. OpenAI also introduced a $500 tier. The report estimates that it offers about 21% more Astra than the old $200 plan, but less Sol-class API-equivalent value. It identifies the advertised 300-token-per-second Ultrafast mode as a potential draw, while saying testing is ongoing.
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 Limits Shape the Economics
The estimated value gap matters because it raises a question about who pays for heavy subscription use. SemiAnalysis estimates that subscriptions account for about 10% of Anthropic revenue but can 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 revenue, though it does not provide a comparable figure in the supplied material.
SemiAnalysis’s margin estimates illustrate how much usage patterns matter. Assuming a subscriber uses the full allowance and API gross margins are 92%, it estimates gross margins of about minus 369% for maxed-out Opus 5.5 use and about 1% for Fable 5.1. At 20% average utilization, its estimates rise to about 6% and 80%, respectively. These are modeled outcomes based on the report’s assumptions, not disclosed company accounts. The report’s analysis suggests that the most expensive model usage drives much of the subsidy, while older or cheaper models can cost less to serve.
For customers, the practical value depends on which models they use, how often they use them and whether they hit plan limits. OpenAI’s Pro plans have no five-hour usage window, according to the report, which could help people with bursts of intensive work use more of their monthly allowance. That feature may narrow the real-world difference for some users, even though SemiAnalysis says it does not erase the estimated API-value gap.
As an affiliate, we earn on qualifying purchases.
Recent Price Cuts and Allowances
The report tracks price reductions alongside changes in subscription limits. Anthropic cut Fable 5.1 cache-read prices by 75% compared with Fable 5, and cut Opus 5.5 input and output prices by 20% and cache-read prices by 60% compared with Opus 5. SemiAnalysis says Fable’s token limits did not rise when version 5.1 shipped. Opus allowances rose by about 20% on Max and 50% on Pro, but the report estimates that those increases did not fully offset the price cuts in API-equivalent terms.
SemiAnalysis says OpenAI made no corresponding limit increase when GPT-6.1 Sol shipped, and estimates that the change reduced API-equivalent value on the $200 plan by about 30%. It also says OpenAI’s Pro 100, 200 and 500 tiers now return identical tokens per dollar, and that the company removed “5x more usage” and “20x more usage” multipliers from its pricing page. Those estimates describe the plans as measured in the report; future prices and allowances may differ.
As an affiliate, we earn on qualifying purchases.
Usage and Future Limits Remain Unclear
The API-equivalent figures are estimates tied to a specific workload, model versions, list prices and measured plan limits. The coding-agent example is dominated by cached input, so customers with different mixes of fresh input, cache writes and output could see different results. The figures do not establish how much a typical subscriber uses, what each company’s actual per-customer serving costs are, or whether subscribers receive the full measured allowance in every circumstance.
The report’s margin calculations also rely on assumptions, including 92% API gross margins and selected utilization rates. The supplied material does not include audited subscription profitability or a full breakdown of OpenAI’s subscription share of revenue. It is also unclear how long the measured limits and prices will remain in place, how customers will use the new $500 tier, or how Ultrafast mode performs in practice.
As an affiliate, we earn on qualifying purchases.
Plan Changes Will Test the Estimates
The next concrete date identified in the report is October 29, when existing $200 ChatGPT subscribers are due to lose their grandfathered limits under the described change. SemiAnalysis says it is still testing OpenAI’s Ultrafast mode. Future comparisons will depend on whether OpenAI or Anthropic changes prices, token allowances or model access again, and on how subscribers use the available models. Until those details emerge, the five-to-six-times estimate is best read as a snapshot of specific plans and a defined workload.
As an affiliate, we earn on qualifying purchases.
Key Questions
What does the five-to-six-times estimate compare?
It compares the estimated API list-price value of measured plan allowances for Claude Opus 5.5 and GPT-6.1 Sol on a coding-agent workload. It is not a claim that every subscriber gets that much more practical value.
Why might the estimate change?
It depends on token allowances, model prices, model choice and the mix of cached input, fresh input, cache writes and output. Providers can also change plan limits and prices.
What changed for OpenAI’s $200 plan?
SemiAnalysis says OpenAI roughly halved token allowances across model tiers. Existing subscribers keep the previous limits until October 29; new purchases get the lower limits immediately.
Does the report show that Anthropic loses money on every subscriber?
No. Its margin figures are modeled estimates under specified usage and API-margin assumptions. SemiAnalysis estimates very different outcomes at full usage and at 20% average utilization.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
