ChatGPT Pro $500 Plan: Pro Tiers, Plus Limits, Ultrafast (Oct 2026)
The short version: according to OpenAI's Codex pricing docs, ChatGPT Pro now starts at $100 per month and comes in three tiers: $100, $200, and $500. Only the ChatGPT Pro $500 tier gets Ultrafast mode, as stated in the October 8 changelog. Pro currently has no five-hour limit; check the official page for weekly limits. If you are on Plus today, first work out whether you are hitting the five-hour window or the weekly allowance, then decide between upgrading and buying credits.
Everything below is current as of October 2026 and comes from OpenAI's pricing page and changelog. Prices and limits change often, so confirm on the official pages before you pay.
What happened on October 8?
The October 8 changelog entry is titled "GPT-6.1 Sol Ultrafast arrives in Codex and ChatGPT Work." OpenAI says Ultrafast mode speeds up token generation, and it is available on the Pro $500 tier and on eligible Enterprise and Edu plans. It is off by default for Enterprise and must be turned on by a workspace owner. Ultrafast for this model supports inference residency in the US and Europe (EEA plus Switzerland).
The pricing page also lists Pro as $100, $200, or $500 per month. I did not find specific numbers on the page for how the allowances differ between the three tiers, so this article does not invent any.
How do the plans compare on price?
| Plan | Price | Codex-relevant notes |
|---|---|---|
| Free | $0 | GPT-6 Luna at standard speed in the desktop app, depending on rollout |
| Go | $8/month | Same model access as Free |
| Plus | $20/month | Includes GPT-6.1 Sol and GPT-6 Luna; credits can be purchased separately |
| Pro | $100 / $200 / $500 per month | No five-hour limit at the moment (check the official page for weekly limits); Ultrafast on the $500 tier only |
| Business | $20 per user/month (billed annually, 2-seat minimum), $25 billed monthly | Standard Business uses the same estimate basis as Plus |
| Enterprise / Edu | Contact sales | With flexible pricing there is no fixed rate limit and usage scales with credits; plans without flexible pricing are close to Plus seat limits (for most features) |
| API key | Standard API pricing | No cloud features |
How much can Plus do in a five-hour window?
OpenAI gives estimates, not hard caps, counted as local messages per five-hour window (Plus and standard Business):
| Model | Estimated local messages per five hours |
|---|---|
| GPT-6.1 Sol | 15–160 |
| GPT-6 Sol | 15–150 |
| GPT-6 Astra | 5–45 |
| GPT-6 Luna | 350–3,000 |
The ranges are wide because tasks vary in size. The docs also note that cloud tasks can use more of your allowance. For the full breakdown across Plus, Pro, and Business, see our earlier post, how Codex usage limits work.
Why do Fast and Ultrafast use more of your allowance?
Both modes trade allowance for speed. The multipliers are relative to standard speed on the same model:
| Mode | Allowance use within a subscription | Purchased credits / Enterprise pay-as-you-go |
|---|---|---|
| Fast | 2.5x | 2x |
| Ultrafast (GPT-6.1 Sol, GPT-6 Astra) | 8x | 6x |
In other words, running the same batch of tasks on Ultrafast burns subscription allowance roughly 8 times faster. Turn it on only when speed matters, and keep batch or non-urgent work at standard speed. One caveat: the changelog says only that GPT-6.1 Sol Ultrafast is available on the Pro $500 tier, while the multiplier table also lists GPT-6 Astra, so check the official pages for exact availability.
Should you upgrade or buy credits?
When you run out of allowance you do not have to upgrade first; OpenAI lets you buy credits to keep going. Work through this in order:
- Check when you hit the limit. If heavy daily use gets you blocked by the five-hour window, the window is your bottleneck.
- If the window is enough but you run dry in the second half of the week, the weekly allowance may be the bottleneck. See the official page for how it is counted.
- Hitting limits only occasionally (say once or twice a month): credits are the better deal, and you avoid paying a higher monthly fee all the time.
- Hitting limits in almost every window: Pro has no five-hour limit, so it is the right fit. Start with the $100 tier and see whether you need to go higher.
- If the only thing you want is more speed and the budget allows it: that is when the $500 tier's Ultrafast is worth a look.
- For teams: look at Business per-seat pricing, and at Enterprise if you need SCIM, compliance, and similar capabilities.
Standard-speed credit pricing is also on the pricing page. For example, GPT-6.1 Sol is 50 credits per million input tokens, 2.5 per million cached input tokens, and 250 per million output tokens. The pricing page's typical task cost (2–15 credits) refers to a different model and gives no matching figure for GPT-6.1 Sol, so I do not quote it here.
How do I track allowance resets and outages?
Allowance resets and service outages often get mixed up. Publicly documented resets are in our Codex usage limit reset timeline, and the capacity errors seen in October are covered in Codex errors and troubleshooting. To compare with another vendor's subscription terms, see Claude Max 5x vs. 20x.
Pricing pages and changelogs get edited. The cdnping site changes tracker shows line-by-line additions and removals for pages from OpenAI, Codex, and others. When Codex seems unstable, check Is Codex down? and OpenAI API status first to tell a service problem from your own allowance.
What if you would rather pay monthly in RMB?
AGI Plan is a commercial partner of cdnping. It is a monthly-subscription gateway to multiple model providers, billed in Chinese yuan (RMB), with four tiers (Lite, Plus, Pro, Max) that differ only in allowance. It is not an officially authorized OpenAI channel and is not the same thing as the ChatGPT plans above. See the AGI Plan pricing page for available models and prices.
Sources
- OpenAI Codex pricing docs, accessed October 2026
- OpenAI changelog (October 8 entry), accessed October 2026
