Free guide
Codex Ultrafast on Pro $200: Same Model, Faster Tokens, 6x the Usage
OpenAI brought Ultrafast to the $200 Pro plan in Codex, starting with GPT-6.2 Sol. It doesn't change the model. It writes the same model's tokens faster, and it spends your allowance faster too. Here's how it works, what it costs, and how to switch it on and off.

The short version
- Same model, not a smarter one: OpenAI's docs say Ultrafast generates tokens faster than Standard on the same model, up to 4.5x on GPT-6.2 Astra and Sol.
- The catch is usage: Ultrafast on GPT-6.2 uses 6x your included allowance compared with Standard. Purchased credits cost 3x until February 28, 2027, then 6x.
- Only the writing speeds up. Tests, builds and other tool runs take as long as they did, so a whole task won't be 4.5x faster.
How it actually works
A model writes like a loom weaves
A model writes its answer one token at a time, the way a Jacquard loom weaves cloth one row at a time from its punch cards. The cards decide the pattern. In this picture the cards are the model: what it knows and how it answers.
Ultrafast runs the same cards on a faster loom
OpenAI's Codex docs say Fast and Ultrafast generate tokens faster than Standard on the same model. On GPT-6.2 Astra and GPT-6.2 Sol, Ultrafast generates tokens at up to 4.5 times the speed of Standard. Same cards, same pattern, rows come off faster. OpenAI doesn't document how GPT-6.2 Ultrafast is served, so we don't guess at the hardware.
Every row costs more thread
Speed isn't free. Codex's pricing page lists speed multipliers relative to Standard for the same model: Ultrafast on GPT-6.2 Astra and Sol uses 6x your included usage. Fast uses 2x. Your prompt, files, history, tool results and the reply all count as tokens, so the whole turn is billed at the higher rate, not just the writing.
The rest of the job keeps its own pace
Ultrafast speeds up token generation. Your tests still take as long to run, your builds still take as long to compile, and any time spent waiting on tools stays the same. That's why a task that's mostly running tests gains less than one that's mostly writing code or text.
How to set it up
1.Check that you have it
On ChatGPT Pro $200, Ultrafast works with GPT-6.2 Sol. Pro $500 can use it with GPT-6.2 Astra or Sol. Enterprise and Edu workspaces need an owner to turn Ultrafast on. Pick GPT-6.2 Sol first in Codex.
/model2.Turn it on in the Codex CLI
In the CLI, type / to see the commands for your model, then type /ultrafast. Type it again to return to Standard. The CLI saves your selection, so check which speed you're on before a long run.
/ultrafast # type it again to go back to Standard /ultrafast3.Or use the speed menu in the app
In the Codex desktop app or ChatGPT Work on the web, open the model or effort picker at the bottom of the composer, select the lightning bolt to open the speed menu, then choose Ultrafast or Standard.
4.Use it where you're actually waiting
Ultrafast pays off when you're watching it work: quick fixes, reviews, back-and-forth edits. For long background jobs that you aren't watching, Standard does the same work for a sixth of the included usage.
5.Building on the API? Set the service tier
In the Responses API, set the model to gpt-6.2-sol or gpt-6.2-astra and set service_tier to ultrafast. OpenAI recommends WebSockets for agents that make many tool calls, because network overhead can eat the latency gains. API Ultrafast costs 3x the Standard rate through February 28, 2027, and 6x from March 1, 2027.
model: "gpt-6.2-sol" service_tier: "ultrafast"
The fine print
- Speeds are 'up to' figures from OpenAI's docs. Real speed depends on the task and on load.
- Included usage on Ultrafast is 6x Standard for GPT-6.2 Astra and Sol. Purchased credits and eligible pay-as-you-go usage cost 3x through February 28, 2027 and 6x from March 1, 2027.
- Pro $200 gets Ultrafast with GPT-6.2 Sol only. Pro $500 gets both GPT-6.2 models.
- For GPT-6.2 Astra and Sol, Ultrafast supports inference residency only in the United States.
- If you sign in to Codex with an API key, you pay API token rates, not ChatGPT allowance multipliers.
- Codex and ChatGPT Work share pricing, credits and usage limits.
FAQ
Does Ultrafast make GPT-6.2 Sol smarter?
No. OpenAI's docs say Ultrafast generates tokens faster than Standard on the same model. It changes the speed, not the model.
How much faster is Ultrafast in Codex?
On GPT-6.2 Astra and Sol, up to 4.5 times the token speed of Standard. Whole tasks gain less, because tests, builds and other tool runs don't speed up.
How much usage does Ultrafast use?
6x your included usage compared with Standard on GPT-6.2 Astra and Sol. Purchased credits are 3x until February 28, 2027, then 6x.
Is Ultrafast on the $200 Pro plan?
Yes, with GPT-6.2 Sol. OpenAI announced it on November 17, 2026. Pro $500 can use it with GPT-6.2 Astra or Sol.
How do I turn Ultrafast off in the Codex CLI?
Type /ultrafast again. Repeating the active speed command returns you to Standard, and the CLI saves your choice.
Sources
- OpenAI on X: GPT-6.2 Astra and Sol, Ultrafast on Pro 200
- Codex docs: Fast and Ultrafast modes
- Codex pricing: speed multipliers
- Codex what's new: GPT-6.2 Astra and Sol
- OpenAI API: Ultrafast mode
- OpenAI API: GPT-6.2 Sol model page
2026-11-18 23:15 PT against OpenAI's Codex speed, pricing and what's-new docs, the API Ultrafast guide and GPT-6.2 Sol model page, and @OpenAI's November 17 post.
Want help building with tools like this?
Join the Shipping Skool community free. Paid members get live bootcamp calls every week and weekly-updated courses.
Join Shipping SkoolMore free guides
Chrome DevTools MCP: Let Your Coding Agent See the Page It Builds
Google's free, open-source way to give your coding agent a real Chrome window and the DevTools you already use. What it does, how it works, how to set it up, and the fine print.
GPT-6.2 Drops Its Reasoning When You Edit Old Messages: How to Keep It
Every reasoning item is now tied to the conversation before it. Edit that history and the API drops the reasoning that follows. Here's how it works and how to keep it.
Inkling 2 Mini: Run Thinking Machines' Looped AI Model on Your Mac
Thinking Machines' new open-weights model runs on a 32GB Mac with the wifi off. What it is, how the loop trick works, how to set it up, and the fine print.