GPT-6.2 Ultrafast in Codex: Which Plans and How Fast It Uses Limits (Nov 2026)
Data as of · Official docs checked 18 Nov 2026; launch post: supplied text, live fetch HTTP 403 · Page updated
Short answer (as of 18 Nov 2026): GPT-6.2 Ultrafast: Pro 500 gets both Astra and Sol. OpenAI’s launch wording is “bringing Ultrafast to Pro 200 subscribers, starting with GPT-6.2 Sol”. Eligible Enterprise/Edu workspaces get both once an owner enables Ultrafast. Token generation is up to 4.5x Standard speed; included usage is charged at 6x Standard. Pro uses included allowance first, then available credits.
Cite this: PlanYield, 'GPT-6.2 Ultrafast in Codex: Which Plans and How Fast It Uses Limits (Nov 2026)', checked 18 Nov 2026, https://planyield.dev/guides/gpt-6-2-ultrafast/. CC BY 4.0.
Get a weekly email when these limits change
One email a week: what changed in AI coding plan limits and prices, with sources. No hype, no affiliate links.
By subscribing you agree to the privacy notice. Unsubscribe in one click.
Which plans get GPT-6.2 Ultrafast?
The GPT-6.2 release note, fetched on 18 Nov 2026, confirms Sol access on Pro 200 and both models on Pro 500. Eligible Enterprise and Edu workspaces can use both after owner enablement.
The launch post uses the wording quoted in the short answer. Its supplied saved text was read on 18 Nov 2026; a direct curl fetch returned HTTP 403 that day. The access details above are corroborated by the fetched release note.
The pricing page Pro card still says “Ultrafast mode access on Pro $500”. That card does not yet mention Pro 200; the release note does.
How fast is it?
OpenAI’s speed documentation gives GPT-6.2 Astra and Sol up to 4.5x the token-generation speed of the same model at Standard. This is a token-generation comparison; it does not promise that an entire coding task finishes 4.5x sooner.
How fast does it use the allowance?
For Pro, the speed docs describe a billing order: use the included allowance, then available credits. The pricing table publishes the following rates per model, relative to Standard for that same model. These are model rates, not a separate Pro 200 rate.
| Model | Included usage | Eligible credits / PAYG through 28 Feb 2027 | Eligible credits / PAYG from 1 Mar 2027 |
|---|---|---|---|
| GPT-6.2 Astra | 6x Standard | 3x Standard | 6x Standard |
| GPT-6.2 Sol | 6x Standard | 3x Standard | 6x Standard |
The temporary credit rate applies to eligible purchased-credit overages and Enterprise pay-as-you-go usage. Included usage stays at 6x throughout. These multipliers do not supply a fixed number of coding tasks or hours.
API-key billing has its own rates
GPT-6.2 Astra/Sol Ultrafast API pricing: 3x the Standard rate through 28 Feb 2027, then 6x starting 1 Mar 2027 (Pacific time). API-key billing is separate from ChatGPT included usage. OpenAI API pricing, read 18 Nov 2026.
The speed docs distinguish sign-in methods: Codex with an API key uses API token billing; ChatGPT credit multipliers apply to ChatGPT sign-in.
Turn it on in Codex
In the app, open the model or effort picker, then the lightning-bolt speed menu and select Ultrafast if available. In the CLI, use /ultrafast; run the active command again to return to Standard. The CLI remembers the selection. OpenAI’s speed controls.
Workspace permissions and inference residency
The workspace requirements specify a credit-based or USD usage-based agreement for Enterprise and a credit-based plan for Edu. Owners must enable the models and the separate Ultrafast permission, which starts off. They can grant it to selected users or the whole workspace; per-user spend controls apply.
For GPT-6.2 Astra and Sol, Ultrafast inference residency is US-only. This restriction describes inference residency, not the subscriber’s location.
Sources and verification
- GPT-6.2 release note: fetched with curl and read 18 Nov 2026 (Asia/Singapore).
- Speed and workspace controls: fetched with curl and read 18 Nov 2026 (Asia/Singapore).
- ChatGPT pricing: fetched with curl and read 18 Nov 2026 (Asia/Singapore).
- OpenAI API pricing: fetched with curl and read 18 Nov 2026 (Asia/Singapore).
- Launch post: supplied saved text read 18 Nov 2026; live curl returned HTTP 403. Its exact Pro 200 wording is preserved above.
More: Codex limits · ChatGPT Pro plans · Codex 5-hour limit · 17 Nov GPT-6.2 change
Dataset licence: CC BY 4.0.
FAQ
Which plans get GPT-6.2 Ultrafast in Codex?
Pro 500 supports Astra and Sol; Pro 200 supports Sol. Eligible Enterprise and Edu workspaces support both once an owner enables the models and Ultrafast.
How much faster is GPT-6.2 Ultrafast in Codex?
OpenAI publishes up to 4.5x Standard token-generation speed for the same GPT-6.2 model. It is not a whole-task completion-time guarantee.
How fast does Codex Ultrafast use included allowance and credits?
Pro uses included allowance first, then available credits. GPT-6.2 Astra and Sol each use included allowance at 6x Standard throughout. Eligible purchased-credit overages and Enterprise PAYG cost 3x through 28 Feb 2027, then 6x from 1 Mar 2027 (Pacific time).
What does GPT-6.2 Ultrafast in Codex cost with an API key?
API-key usage is billed separately at API token rates. For GPT-6.2 Astra and Sol, Ultrafast costs 3x Standard through 28 Feb 2027 and 6x from 1 Mar 2027 (Pacific time).
How do I enable Codex Ultrafast, and where can inference reside?
Select Ultrafast in the Codex app’s speed menu or use /ultrafast in the CLI. Workspace owners must enable model access and the separate Ultrafast permission. GPT-6.2 Ultrafast inference residency is limited to the United States.