OpenAI Drops 5x/20x Usage Guarantees, Adds $500/Month Pro Max Tier

On September 27, users observed that OpenAI quietly revised its ChatGPT pricing page. The two existing Pro tiers — previously $100/month and $200/month — were renamed to "Pro Standard" and "Pro More" . The explicit quantitative promises "5x more usage than Plus" and "20x more usage than Plus" were removed entirely, replaced by the ambiguous phrase "More usage than Plus."

Previously, the $100 tier guaranteed 5× the usage allowance of the $20 Plus plan, and the $200 tier guaranteed 20×. These numbers served as the primary decision criteria for subscribers. Their disappearance creates a transparency gap: OpenAI's own documentation still retains the old multiplier language, causing a discrepancy between the pricing page and the docs.

GitHub PR Reveals Three‑Tier Structure

On September 25, the openai/codex repository merged pull request #47971 titled "Add Pro Max plan support and update Pro display names." The 24‑file change discloses three key points:

A new promax plan type with the display name "Pro (Max)" .

Internal renaming of existing tiers: prolite → display name "Pro" (i.e., Pro Standard), and pro → display name "Pro (More)" .

No prices are hard‑coded in the PR, but earlier leaked in‑app strings indicate Pro Max will cost $500/month and is described as "Fastest Work and Codex" .

The removal of the 5×/20× multipliers likely stems from the need for a new descriptive framework that can accommodate three distinct Pro tiers without locking the company into specific quantitative ratios.

Pro Max Speculation: Speed and Hardware

The leaked "Fastest Work and Codex" description points to two differentiators: the highest inference speed and full access to the Codex programming agent. Unverified reports suggest Pro Max may run on Cerebras inference hardware to deliver significantly faster response times, though OpenAI has not confirmed this.

Product Logic: Compute Tiering Amid Capacity Constraints

The article infers a clear product rationale: the compute demands of a future GPT‑6 Astra model are so large that the $200 tier had to pause registrations. Introducing a $500 "ceiling" plan allows OpenAI to allocate premium compute resources to the highest‑paying users while protecting the experience of $100 and $200 subscribers from resource contention.

Community Reaction: Polarized

Reactions split into two camps:

Supporters argue that tiered pricing reflects real inference costs; stronger models consume more compute, so a high‑end plan is reasonable. For teams replacing human labor, $500/month may still be cost‑effective. Some even speculate the restructuring precedes price reductions.

Critics focus on transparency: replacing concrete usage multipliers with vague language is widely seen as a negative signal. Social media shows accusations of "harvesting users" (割韭菜) and calls to unsubscribe.

The core concern: by erasing quantitative commitments while launching a more expensive tier, OpenAI gains broad flexibility to adjust allowances downward for existing Pro users — a potential "hidden downgrade."