OpenAI ships Pro 500, halves Pro 200 usage Oct 30

The seat you budgeted for in August is not the seat you get in November. At DevDay 2026 on September 29, OpenAI launched Pro 500 at $500 a month — and in the same breath cut the usage allowance on the $200 Pro 200 plan roughly in half, effective October 30. If your team runs Codex or ChatGPT Work on Pro 200, your OpenAI pricing just moved, and not in your favor.

What actually happened

Pro 500 is $500/month and carries 25× the Plus usage allowance, plus exclusive access to Ultrafast — OpenAI's premium speed tier, which The Next Web reports hits up to 300 tokens per second in Codex, about eight times standard. Sam Altman priced Ultrafast on stage at roughly six times the standard token rate. Every Pro subscriber also gets one Dot, OpenAI's new always-on background agent, whose chats sit outside the normal message limits. Dots are not available at launch to Pro users in the EEA, Switzerland, or the UK.

The other half of the announcement is the part that costs you money. From October 30, Pro 200's included usage in ChatGPT Work and Codex drops from 20× to 10× the Plus allowance, and weekly GPT-6 Pro messages in chat fall from 200 to 100. Existing subscribers keep their current limits through October 29 and get a one-time grant of usage credits worth $2,500 — which expire at the end of the year. OpenAI did commit to not bringing back Pro 200's old five-hour usage window.

Read that sequence plainly: the $200 tier lost half its headroom, the replacement costs $500, and the sweetener is a credit that evaporates on December 31.

Why AI seat pricing matters for your business

The credit grant is a timing trap. $2,500 in credits that expire at year-end is not compensation, it's a nudge to burn usage in Q4 and normalize a higher run rate before renewal. If you take it, spend it on work you already planned — a backlog of migrations, a batch of test generation — not on new habits your January budget can't carry.

Price your seats per outcome, not per subscription. Ten seats at $500 is $60,000 a year. That is a hire. Before you upgrade anybody, measure what each seat actually produced last quarter: PRs merged, tickets closed, hours returned. Most teams we audit find two or three people genuinely pinned against the ceiling and six who would not notice a downgrade to Plus.

Halved limits are a repeat event, not an incident. OpenAI paused Pro 200 signups in September, reopened them on stage today, and cut the plan's allowance in the same keynote. Treat quota as a variable your architecture accounts for. Keep model access behind one internal interface, log per-task cost and token spend, and make swapping providers a config change instead of a refactor. When a tier gets repriced, you should be able to move the overnight batch job to a cheaper model in an afternoon.

Sources: The Next Web, Engadget.