ChatGPT Pro 500 for Codex: Pro 100, 200 and 500 compared
This research library uses AI-assisted source research and drafting. Linked sources support product claims; analysis and proposed exercises are our interpretation. Unless an article documents a test and its results, do not read it as a hands-on review or an independently verified benchmark.
ChatGPT Pro 500 is the new top Pro tier OpenAI launched at DevDay on 29 September 2026. It costs $500 a month, gives 25 times the ChatGPT Plus usage allowance, and is the only Pro plan that includes Ultrafast in Codex and ChatGPT Work. Pro 100 and Pro 200 remain on sale without Ultrafast, as the Codex speed docs and the Pro tiers help article describe.
ChatGPT Pro tiers side by side
This table uses only what OpenAI has published today. Where a number is missing, the source does not state it.
| Plan | Price per month | Ultrafast in Codex | Usage allowance | Notes |
|---|---|---|---|---|
| Plus | $20 | No | Baseline | Codex on web, CLI, IDE extension and iOS |
| Pro 100 | $100 | No | Higher than Plus, exact multiple not published | Credits can be bought, but do not add Ultrafast |
| Pro 200 | $200 | No | Lower for new subscribers than for grandfathered ones | Existing eligible subscribers keep their previous allowance through 29 October 2026 |
| Pro 500 | $500 | Yes | Highest, 25 times the Plus allowance | Ultrafast uses included usage first, then credits |
All Pro tiers include everything in Plus. ChatGPT Work and Codex share one pool, with the same pricing, credits and usage limits.
Pro 500 vs Pro 200: what actually differs for Codex
Two things separate the tiers for a Codex user: Ultrafast access and allowance size.
Ultrafast is a speed tier, not a model. GPT-6 Astra Ultrafast generates tokens up to 8 times faster than GPT-6 Astra in Standard mode in Codex, around 300 tokens per second according to the DevDay recap. OpenAI notes that this measures token generation speed, not total task completion time. GPT-6.1 Sol Ultrafast is announced as coming soon.
Speed costs allowance. For GPT-6 Astra, Ultrafast draws your included subscription limits at 8 times the Standard rate. Once the included allowance runs out, it bills purchased credits at 6 times the Standard rate. So the 25 times Plus allowance on Pro 500 shrinks fast if Ultrafast is your default.
On Pro 500 you select Ultrafast in the model picker. On Pro 100 and Pro 200 it does not appear, and the docs are explicit that buying credits does not add it.
How credits behave on each Pro tier
Credits extend usage after your plan allowance runs out. They cover Codex, ChatGPT Work, and ChatGPT for Word, Excel and PowerPoint.
The Codex pricing page lists credit rates per million tokens:
| Model | Input credits per 1M tokens | Output credits per 1M tokens |
|---|---|---|
| GPT-6 Astra | 250 | 1,250 |
| GPT-6.1 Sol | 50 | 250 |
| GPT-6 Luna | 2.5 | 12.5 |
Speed modes multiply those rates. Fast mode uses included limits at 2.5 times the Standard rate and credits at 2 times. Ultrafast on GPT-6 Astra uses included limits at 8 times and credits at 6 times. The billing multipliers do not describe speed gains.
One exception matters for teams with API accounts. If you run Codex with an API key, it uses API token pricing, and the ChatGPT credit multipliers do not apply.
Set Fast mode as your Codex default
Fast mode is the middle option, and it works on every tier where the model supports it. GPT-6.1 Sol, GPT-6 Astra, GPT-6 Sol and GPT-6 Luna support Fast mode where available. It runs in the ChatGPT desktop app, the Codex CLI and the IDE extension when you sign in with ChatGPT.
In the CLI, /fast toggles it and /statusline shows it in the footer. To make it the default, the docs name two settings in config.toml:
service_tier = "fast"
[features]
fast_mode = true
A practical Pro 500 workflow in Codex looks like this. Keep GPT-6.1 Sol on Standard or Fast for long runs you do not watch, such as a test migration or a refactor across many files. Switch to GPT-6 Astra Ultrafast in the model picker for short interactive loops where you sit and wait for output. Switch back when the loop ends, because Ultrafast work draws your allowance at eight times the Standard rate.
Keep your repo instructions in AGENTS.md so a model or speed switch does not change what Codex knows about the project. We use the same rule in our review methodology: the plan you pick changes speed, not the review you owe the diff.
Should Pro 200 subscribers move before 29 October 2026?
Existing eligible Pro 200 subscribers keep their previous allowance through 29 October 2026. New Pro 200 subscriptions start with a lower allowance than grandfathered ones. OpenAI's help text does not say what the grandfathered allowance becomes after that date, so check your plan page before then.
A simple way to decide:
- Stay on Pro 200 if you never hit your limit and do not need Ultrafast.
- Drop to Pro 100 if Pro 200 already leaves most of your allowance unused.
- Move to Pro 500 if you want Ultrafast, or if you regularly buy credits on Pro 200 to finish the week.
- Use an API key for bursty automation where token pricing is easier to budget than a subscription.
One more date affects plan choice. GPT-5.5 retires from ChatGPT, ChatGPT Work and Codex on all plans on 14 October 2026. The OpenAI API is not affected.
Ultrafast outside Pro 500
Ultrafast is also available on eligible Enterprise and Edu plans. For Enterprise workspaces it is off by default, and workspace owners enable it for selected users or the whole workspace. Other self-serve plans do not get Ultrafast at launch, even with purchased credits. Workspaces that require inference residency outside the United States cannot use Ultrafast.
Before you upgrade, run one normal week on your current tier and note how often Codex stalls on your limit.