ChatGPT Pro 500 and the Pro 200 Cut: What Changed

On September 29, 2026, OpenAI launched a $500 ChatGPT Pro plan and lowered the usage included in the $200 plan, without changing its price. New Pro 200 subscribers get the smaller allowance now. Existing ones keep their current allowance through October 29, then move to the smaller one. If you run Codex or ChatGPT Work for hours a day, this changes your budget. It follows a June change to how ChatGPT's memory works.

What changed in ChatGPT Pro, in one sentence?

On September 29, 2026, OpenAI added a $500 Pro tier and cut the usage included in the $200 tier, keeping its price.

OpenAI's help page on Pro tiers says new Pro 200 subscriptions "include a lower usage allowance than previously offered". It gives the reason as reflecting "our increasingly efficient models". The page does not publish the new numbers. According to The Next Web, Pro 200's allowance in ChatGPT Work and Codex drops from 20 times the Plus allowance to 10 times. New subscribers get the lower figure now, and grandfathered subscribers get it from October 30.

How do the three Pro plans compare?

Ultrafast is a faster mode of GPT-6 Astra. The Next Web reports it generates up to 300 tokens per second in Codex. It also burns allowance fast. OpenAI's help page says Ultrafast uses your included allowance at 8 times the Standard rate, and Fast mode at 2.5 times. OpenAI's help page adds that buying credits on Pro 100 or Pro 200 does not unlock Ultrafast, at least at launch.

What happens to existing Pro 200 subscribers?

You are grandfathered if you had an active Pro 200 subscription at any point from September 22 to 10 a.m. Pacific Time on September 29, 2026. Grandfathered accounts keep the previous allowance through October 29, 2026. After that, the subscription moves to the lower allowance at the same $200 price. OpenAI says only affected users get an email.

The Next Web adds two details that OpenAI's help page does not spell out. Existing subscribers get a one-time grant of usage credits worth $2,500, which expires at the end of the year. GPT-6 Pro messages in chat fall from 200 to 100 a week.

Why does Pro usage run out faster than you expect?

OpenAI's help page lists what drives consumption: the model, the amount of context processed, reasoning, tool use, retrieval, caching and speed mode. Two similar tasks can cost very different amounts. A long Codex session that rereads a large repository with Fast mode on (2.5 times the Standard rate) can use far more allowance than a short chat.

A few habits stretch a smaller allowance. Keep Standard mode for routine work and save Fast mode for the tasks where waiting costs you. Start a new session when the history gets long. And check Settings, then Usage, before the October 29 cutover, so you know your real weekly burn.

Has ChatGPT changed for existing users before?

Yes. On June 4, 2026, OpenAI began rolling out a new memory system built on what it calls dreaming. Before, saved memories were a list of items you could read and edit. The new system synthesizes memory in the background from your whole chat history, and you review it through a memory summary. OpenAI's memory help page says that summary "does not necessarily include every detail or source that Memory can reference."

At least one user noticed. A post on OpenAI's community forum from August 18 says ChatGPT stopped following persistent instructions after the June rollout. The poster writes that work "now requires repeated correction". OpenAI Support replied by suggesting the legacy option. That option still exists: in Settings, then Memory, you can select Saved memories to go back to the explicit list.

Read the fine print

OpenAI's case for each change is reasonable on its own. Its models got more efficient, so a smaller allowance may do as much work as before. OpenAI says background memory fixes saved memories that went stale, and it calls the new system significantly more capable. It also kept a way back for people who want the old list.

Subscribers see it differently. One reply on a forum thread about Pro 200 calls it "Same price, half the product". The halving that The Next Web reports applies to the Work and Codex allowance, not the whole plan. Our read is that both sides have a point. Both changes show the same thing: the vendor decides what a plan includes and how memory works inside it, and can change either one.

The bottom line

If you lean on Codex or ChatGPT Work with Pro 200, measure your usage now. Compare it with the allowance you will have after October 29. Pro 500 costs $300 more a month. Credits cover occasional overflow in Codex and Work. Moving part of the work to another tool saves money but costs setup time. Whichever you pick, keep the context you depend on somewhere you control.

How do you keep your working memory when a plan changes?

A long Codex or Work session teaches the agent a lot about your project: its conventions, the commands that work, the mistakes it should not repeat. ChatGPT keeps some of that in a memory summary that OpenAI synthesizes, and its help page says the summary may leave out details that Memory can still use. If a plan change pushes part of your work to Claude Code or Cursor, that context stays behind. You end up explaining your project again in each tool.

memU keeps that knowledge in a shared wiki of readable Markdown files. It works with ChatGPT in Work mode and with Codex. Claude Code and Cursor are supported too, among other agents. ChatGPT's regular chat mode is not supported. A background task picks up new sessions, and the agent itself decides what to keep and writes it down as memory or skill Markdown. memU makes no LLM calls of its own. It stores, embeds and retrieves what the agent wrote, so a different host can retrieve it later.

To try the hosted version, get an API key from memu.so and send your agent one message:

Read https://memu.pro/SKILL.md, follow its instructions to install and configure memU, API Key is YOUR_MEMU_API_KEY.

The support matrix for each agent and the self-hosted setup are in the memU GitHub repository.

FAQ

How much usage does ChatGPT Pro 200 include now?

OpenAI's help page says new Pro 200 subscriptions get a lower allowance than before but does not publish the figure. The Next Web reports that the Work and Codex allowance falls from 20 times Plus to 10 times from October 30.

When does the Pro 200 change apply to existing subscribers?

If you had an active Pro 200 subscription between September 22 and 10 a.m. Pacific Time on September 29, 2026, you keep the previous allowance through October 29, 2026. The lower allowance applies after that, at the same $200 price.

What is ChatGPT Pro 500?

A $500 monthly plan with the highest included usage of the three Pro plans and the only one with Ultrafast, a faster mode for GPT-6 Astra. Ultrafast uses the included allowance at 8 times the Standard rate.

Can I switch ChatGPT back to the old saved memories?

Yes. In Settings, then Memory, select Saved memories to use the legacy list of individual memory items. Select Try improved memory to return to the new system.