Pricing guideUpdated
Every ChatGPT plan that comes with Codex, how its usage limits and credits work, and what coding costs past them, with Fabricate's prices beside them.
Codex comes with ChatGPT plans. Free, and Go at $8 a month, include it in the desktop app on GPT-6 Luna as it rolls out. Plus, at $20, adds the CLI, your editor, the web and cloud tasks, and Pro costs $100, $200 or $500 a month. Business is $25 a user a month, or $20 billed yearly.
One app, one month
About 290 credits
Pro at $25 covers it
350 credits a monthEnough for about 15 new apps or 40 changes
$8a month for Go
$25a month for 350 credits
Billed monthly, as listed on the pricing page in OpenAI's Codex docs. Each but the API key is a ChatGPT plan, so it pays for ChatGPT as well. Business costs $20 a user a month billed yearly, for 2 users or more.
$0
For quick coding tasks, on GPT-6 Luna at standard speed
Codex in the ChatGPT desktop app, as it rolls out. The pricing page lists the CLI, the editor extension and cloud tasks from Plus.
$8a month
For lightweight coding tasks, on GPT-6 Luna at standard speed
Codex in the ChatGPT desktop app, as it rolls out, on ChatGPT's cheapest paid plan.
$20a month
About 15 to 160 local messages every five hours on GPT-6.1 Sol, or 350 to 3,000 on GPT-6 Luna, and weekly limits may apply
Codex in the CLI, your editor, the web and on iOS, cloud tasks, GitHub code review and Slack, with GPT-6.1 Sol. ChatGPT credits extend it past the limits.
$100a month
No five-hour limit for now, in three sizes, $100, $200 and $500 a month, picked by how much you use
Everything in Plus. Pro $500 adds Ultrafast mode, which runs GPT-6 Astra and GPT-6.1 Sol faster for 8x the usage.
$25a user a month
$20 a user billed yearly, from 2 users
A standard seat gets Plus's estimates, and ChatGPT credits extend it past them
A shared workspace with SAML single sign-on and MFA, larger cloud machines, and no training on your business data by default.
Custom
Plus's limits on each seat, or no fixed limits with flexible pricing, billed in credits
Everything in Business, plus SCIM, role-based access, audit logs, data retention and residency controls, and priority processing.
API rates
Billed per token: GPT-6.1 Sol costs $2 per million tokens in and $10 out
Codex in the CLI, the SDK and your editor, with the models your key can use. No cloud tasks, GitHub code review or Slack.
On ChatGPT's plans, Codex isn't priced by the request. It draws on the plan's usage allowance, which local work and cloud chats share, and which ChatGPT Work, for larger tasks in ChatGPT, uses too. OpenAI gives estimates, not fixed counts: on Plus and a standard Business seat, about 15 to 160 local messages every five hours on GPT-6.1 Sol, 5 to 45 on GPT-6 Astra and 350 to 3,000 on GPT-6 Luna, and weekly limits may apply as well. Cloud tasks can use more than local messages. Pro currently has no five-hour limit.
How fast the allowance goes depends on the model, the size of the task, the context the agent holds and the speed you pick. Fast mode uses it at 2.5x the standard rate and Ultrafast, on Pro $500, at 8x, and image generation uses it 3 to 5x faster on average. Voice in the desktop app costs $0.05 a minute of the same budget. The usage dashboard shows what is left, as does /status in the CLI.
At a limit, a turn already running can finish. After that, Plus and Pro can buy ChatGPT credits and keep going without changing plan, Business and Enterprise workspaces on flexible pricing buy workspace credits, and anyone can run local chats with an API key at API rates. Credits are spent per million tokens: GPT-6.1 Sol costs 50 credits in and 250 out, GPT-6 Luna 2.5 and 12.5, and GPT-6 Astra 250 and 1,250, and what a credit costs depends on the plan. Through the API, GPT-6.1 Sol costs $2 per million tokens in and $10 out, though OpenAI notes that API prices don't measure what a plan includes.
ChatGPT's plans with a monthly price, each beside Fabricate's plan closest in price. Codex meters use in five-hour and weekly limits and Fabricate in credits, so compare what a month buys, not the number on the plan.
A month: $0
Codex
Free · $0
For quick coding tasks, on GPT-6 Luna at standard speed
Fabricate
Free · $0
60 credits a month
Enough for about 1 new app or 6 changes. Build and preview apps.
A month: $0 to $8
Codex
Go · $8
For lightweight coding tasks, on GPT-6 Luna at standard speed
Fabricate
Free · $0
60 credits a month
Enough for about 1 new app or 6 changes. Build and preview apps.
A month: $20 to $25
Codex
Plus · $20
About 15 to 160 local messages every five hours on GPT-6.1 Sol, or 350 to 3,000 on GPT-6 Luna, and weekly limits may apply
Fabricate
Pro · $25
350 credits a month
Enough for about 15 new apps or 40 changes. Any number of apps online, up to 5 of your own domains.
A month: $100
Codex
Pro · $100
No five-hour limit for now, in three sizes, $100, $200 and $500 a month, picked by how much you use
Fabricate
Max · $100
1,750 credits a month
Enough for about 75 new apps or 195 changes. Any number of apps online, up to 20 of your own domains.
Most of a month's credits go on changes, not the first build. Here is an example month on one app, drawn against Pro's 350 credits, and a way to price your own.
An example month: one app and 30 changes
One exception: on Pro and Max, when your app uses AI for its visitors, answering them or reading documents they upload, each use is paid from your credits. Serving the app itself never is.
Estimate your month
New apps: 1, Changes: 30. About 290 credits a month
Pro at $25 covers it
The plan price is the easy part. These four decide what a month really comes to.
CodexAt the limit, a turn already running finishes, then Codex waits for the five-hour window or the week to reset. On Plus and Pro, ChatGPT credits keep it going without a bigger plan, and with an API key you can keep working locally at API rates.
FabricateOn Pro and Max, extra credits start at $10 for 100 and never expire, and apps on a paid plan stay online when credits run out.
CodexCodex writes the code, and the domain comes from where you deploy it. Sites, ChatGPT's hosting for what you build in ChatGPT or Codex, connects a domain you already own but doesn't register one, and custom domains aren't in Enterprise workspaces at launch.
FabricatePro connects up to 5 domains you own and Max up to 20, with the secure padlock set up for you.
CodexSites is included with eligible ChatGPT plans during its public beta, and OpenAI's docs point to its help center for the limits. Deployed anywhere else, the app's traffic is billed by that host.
FabricateNot billed. Serving your published app never touches your credits.
CodexYour app's own calls to OpenAI's models go through the OpenAI API, billed per token apart from the ChatGPT plan: GPT-6.1 Sol costs $2 per million tokens in and $10 out, and GPT-6 Luna $0.10 and $0.50.
FabricateOn Pro and Max, each AI answer your app gives and each document it reads is paid from your credits, and "Where your credits went" lists them for each app.
If you write code, Codex is built for that work. It reads a repository, edits files, runs commands and tests, reviews pull requests on GitHub, and runs cloud tasks that keep going while your computer sleeps, from the CLI, your editor, the desktop app, the web or iOS, and the code stays in your own repository. Plus costs $5 a month less than Fabricate's Pro, and it pays for ChatGPT as well, so if you already pay for ChatGPT, Codex costs nothing more until you reach the limits.
Pro, at $100 to $500 a month, suits a full day of coding, with no five-hour limit for now, and Sites can host a site or web app you build with it on eligible plans while it is in beta. If you work in code, Codex is the better fit. We make Fabricate, so we say this part plainly.
Codex's answers come from its own pages, read October 26, 2026. Fabricate's come from our live plans.
The pages this guide quotes, each with the day it was read.
About this guide
Fabricate wrote this guide, and we make one of the two tools in it. Codex's prices are copied from its own pages with the day we read them; Fabricate's come from the same settings that bill you.
The pages we read about each tool are listed here, each with the day we read it. We compared the tools from their own pricing pages and documentation, not by signing up for them. Spotted something out of date? Write to info@fabricate.build.