The Codex Cost Control Workbook is a downloadable PDF reference built directly from this site's own facts module — every priced model's Standard-tier pricing and specs, the long-context repricing rule explained with a worked example, the cache economics worked through, and the Batch discount and rate-limit tiers laid out, gathered into one document with five original worked scenarios, that you can keep on hand instead of re-deriving from the live site each time. It also lists OpenAI's US list price for each ChatGPT plan as of its build date, with the same caveats the live site prints beside them — a list price is a reference point, not necessarily what you pay.

Delivered digitally

After purchase, the workbook is emailed to you as a PDF. Because it is generated directly from the same facts module the live site reads from, it carries the exact figures that were current on the date it was built, with that date printed on the document itself.

Exactly what's in the download

This is written from the generated file itself, not from a plan — every count below was checked by opening the PDF. It's a 12-page, 9-section workbook: the three-lever framework (which order to apply the long-context cliff, prompt caching and Batch, and why that order matters); a full current Standard-tier pricing table covering every priced model on this site, from the GPT-6 generation and GPT-6.1 Terra back to GPT-5.3 Codex, plus each model's context window, max output and knowledge cutoff, and a table of each model's Fast and Ultrafast price as a multiple of Standard; the long-context repricing threshold explained with a worked before/after cost comparison on the Codex CLI's current default; the prompt-caching break-even derived live from the published multipliers — including the lower figure for the deepest read discount and the higher one for the 12-hour cache lifetime — alongside the cache-invalidation checklist this site's own reference page publishes; the Batch discount, its fixed completion window, and exactly which models offer it; the Build, Launch and Grow usage-tier ladder that replaced Tier 1 to Tier 5, plus a Build-tier RPM/TPM table across every priced model and GPT-5.4 Pro's long-context row; regional pricing, OpenAI's US list price for each ChatGPT plan with its caveats, how Fast and Ultrafast count against a plan's included usage and against credits, and which Codex models each plan offers; and five original worked scenarios — a steady CI pipeline, a long agentic refactor crossing the 272,000-token cliff, caching a large reused system prompt, a nightly Batch classification job, and picking a usage tier for burst traffic on a throttled -pro model — each priced end to end and not published as a single narrative anywhere else on this site.

Who it's for

Anyone budgeting a Codex-heavy workflow — a solo developer sizing a personal project's spend, or a team lead putting together a CI-pipeline or agentic-coding budget before committing to a usage tier. It's also useful simply as an offline reference: the live site is always the most current source, but a PDF you already have open doesn't require a connection, and it's easier to hand to a colleague or attach to an internal budgeting document than a link to a page that updates underneath them.

Keeping it current

Because the workbook is regenerated from the same facts module the site itself reads from, a future edition reflects whatever OpenAI has published as of its own build date rather than whatever was true when you first downloaded it. If a figure you're relying on has since changed, the live reference pages on this site are always the most current version of the same numbers.