A breakdown of Claude Max pricing by developer SataEric turned into a viral dispute with OpenAI: the '20x' multiplier in the $200/month plan, according to his calculations, only works within a 5-hour usage window, while the weekly limit is roughly twice as high as the $100 plan. Thibault Sottiaux, head of Codex and ChatGPT at OpenAI, responded that in Codex, '20X' is tied specifically to the weekly limit, and that there are no 5-hour windows on Pro plans. The incident exposed a major issue: Anthropic's pricing page does not specify which window the multiplier is applied to.

What happened

SataEric (@SataEricUX) published a breakdown of Claude Max pricing labels: the $100/month plan promises '5x Pro limits', the $200/month plan — '20x', making the overpayment look fourfold. According to his calculations, the '20x' multiplier works within a 5-hour usage window, and the weekly limit of the top plan is only roughly twice as high as the $100 plan; the discussion title on Hacker News directly states that the weekly limit of the '20x' plan is '10x'. Thibault Sottiaux (Thibault Sottiaux, @thsottiaux), head of Codex and ChatGPT at OpenAI, publicly responded to the breakdown: he stated that in Codex, both '20X' labels refer specifically to weekly limits, that there are no 5-hour limits on either of OpenAI's Pro plans, and that Pro 20X provides 'quite precisely' 20x the usage volume compared to Plus. Sottiaux's response went viral: about 7.2k likes and over 400 replies.

Context

The confusion grew out of the wording on Anthropic's pricing page: the 'Max 5x' and 'Max 20x' plans are advertised with the phrase '5x or 20x more usage than Pro*', but the base measurement window — 5-hour or weekly — is not named. Meanwhile, the agent subscription limits work on two scales: the short window cuts off burst loads within a day, and the weekly limit sets the total volume that can sustain a continuous agent workflow. Until the unit of measurement (messages, tokens, or compute load) and the reset window are named, any multipliers cease to be a comparable metric, and comparing Claude's '20x' against Codex's '20X' is methodologically incorrect by design. For developers whose coding agent subscriptions have become a basic work resource, precise limit labeling has turned from a footer detail into a question of product trust.

Why this matters for the industry

Competition in coding agent subscriptions between Anthropic and OpenAI has moved to the plane of pricing transparency: a competitor's product head publicly used a breakdown of Claude's limit labeling to contrast OpenAI's model, where '20X' is tied to the weekly limit and there are no 5-hour windows on Pro plans. Limit transparency has become an independent market entry tool: disputes of this kind bring vendors earned media, while vague wording brings reputational costs in the key developer audience. The expected consequence for the industry is clarification of pricing pages: explicit indication of the base window, unit of measurement, and weekly thresholds, as well as the emergence of third-party calculators and real limit trackers. For builders, a product niche opens up: quota consumption monitoring tools, alerts for approaching the weekly ceiling, and quota-aware routing between subscriptions and metered API, as limit architecture becomes a separate product layer on par with model quality.

Why this matters for users

Those choosing between Claude Max 20x at $200/month and Codex Pro 20X cannot compare plans by the number in the name: seemingly identical multipliers are tied to different windows, so actual usage volume can differ by orders of magnitude. The working procedure is simple: before upgrading, take your actual usage for a typical week in the Usage tab and compare it with the weekly limit of the chosen plan, not the multiplier from the plan name. For those running continuous agent sessions, the weekly window will be decisive: 5-hour peaks have time to reset, and the weekly ceiling determines whether the subscription will last for the entire task volume, so the $100/month savings or overpayment should be evaluated by it.

What is still unknown / limitations

SataEric's calculations are a single user's estimate without a published methodology, so the claim of a roughly double weekly limit should be read as a hypothesis to be verified by actual usage in Usage, not as an established fact; the thread title 'weekly limit is 10x' is an interpretation, not confirmed data. Sottiaux's statement about 'quite precisely' 20x volume for Pro 20X compared to Plus comes from an interested party and has not been independently verified. Anthropic has no public methodology for calculating limits with an indication of the base window and unit of measurement, so without instrumental measurement, it is impossible to correctly compare Claude's '20x' and Codex's '20X'.

Sources

Author

Look at AI, editorial team