Is There an OpenAI Codex Certification? What Exists in 2026

Facts checked: 9 October 2026

No — as of October 2026, OpenAI does not have an official, proctored certification for Codex, its coding agent. There's free official training (live bootcamps and self-paced courses with quizzes), and there's a brand-new "OpenAI Certification" program — but that program isn't about Codex or coding at all, and it isn't open to the public yet either. If you've seen a "Codex certification" being sold anywhere, it's coming from a third party with no connection to OpenAI.

What OpenAI actually offers

OpenAI announced its first OpenAI Certifications on December 9, 2025, starting with a course called AI Foundations. It's a hands-on credential completed entirely inside ChatGPT — learners work through scenario-based tasks and get feedback in the product itself, not a timed, question-bank exam. OpenAI built the psychometrics behind it with Coursera, ETS (the organization behind the GRE and TOEFL), and Credly by Pearson, which is a real credentialing pedigree — but two things matter more for anyone evaluating it against Codex specifically.

First, AI Foundations is scoped to general ChatGPT and office-productivity AI use, not coding or Codex. Second, it isn't self-enrollment: access currently runs through employer and public-sector pilots — Walmart, John Deere, Lowe's, Boston Consulting Group, Accenture, and the Office of the Governor of Delaware among them — with OpenAI stating broader public access will expand "through 2026" as it incorporates pilot feedback. An individual engineer, or a company that isn't one of those pilot partners, can't sit this exam today at any price.

Codex has free training — but nothing that certifies you

Separately from the AI Foundations program, OpenAI Academy runs actual Codex-specific training, and it's worth knowing what shape that takes because none of it produces a certification either.

• Codex Bootcamp — a free, three-part live webinar series (101: Agentic Coding, 201: Team Workflows, 301: Advanced Automation) with live demos and Q&A. There's no quiz, no assessment, and no credential attached — you show up, you watch, you leave.
• Self-paced Academy pathways — courses like "Build with AI," which covers Codex alongside the OpenAI API, do include a scored quiz, and passing at 80% or higher earns a digital badge issued through Accredible. That's a real step up from a plain completion certificate — there's at least a pass bar — but it's still an open-book, self-administered quiz on general course content, not a proctored assessment of whether you can run Codex against a real, messy codebase.

This is the same shape Anthropic's free Skilljar courses take for Claude Code, which we covered in our piece on the Claude Code certification landscape: good, free, official training that ends in a badge nobody proctored, rather than a credential that verifies skill.

Nobody at OpenAI has answered the question directly

It isn't just that a Codex certification doesn't exist — it's that the question is openly unresolved even among OpenAI's own developer community. A thread on the OpenAI Developer Community forum asked outright whether OpenAI planned to pair Codex Bootcamp with a real, proctored exam. No OpenAI staff member answered. That's a useful data point on its own: for a coding agent OpenAI is actively pushing into CI pipelines and enterprise workflows, there's no public roadmap — official or unofficial — toward a credential that verifies who's actually good at using it.

That gap has already created an opening for unaffiliated training vendors to sell their own "Codex certifications" — course providers with names like DevCerts and igmGuru advertise Codex certificates with no OpenAI backing whatsoever. If you see "Codex certified" on a resume or a training vendor's page, check who actually issued it before assuming it means anything close to what an OpenAI credential would.

A different shape of gap than Claude or Copilot

We've now looked at three vendors and found three different problems. Anthropic has an exam that actually names Claude Code in its syllabus, but it's gated to Claude Partner Network member organizations. GitHub's Copilot exam is open to anyone, but it tests knowledge of the tool rather than judgment on real code. OpenAI is the odd one out: there's no Codex-specific credential at all, official or gated, and the closest thing to an "OpenAI certification" is about general ChatGPT use in the office — not about writing, reviewing, or shipping code with an agent.

In every case, though, the practical answer to the question companies actually care about is the same: none of these programs tell you whether a specific engineer on your team can direct Codex against your repositories, catch a bad diff, or know when a task is a poor fit for an agent to run unsupervised.

What to test for instead

If you're evaluating Codex for your engineering team, what you need isn't a credential about the OpenAI platform in general — it's evidence that specific people can run Codex well against your own code, in the environments it actually runs in: CI pipelines, local terminals, and pre-merge checks, where the failure modes look nothing like a chat-based assistant's.

That's the model behind our own Codex certification track, which focuses on OpenAI's coding agent in CI and local flows — environment setup and sandboxing, reading a failed agent run and deciding whether to retry, narrow the task, or take it back manually, and a proctored, in-IDE assessment where the candidate has to get a failing pipeline green using Codex without weakening the tests to pass. It's built across two levels, Foundations through Practitioner, delivered as a cohort plus async work. See the full list of published tracks if you're comparing Codex against Claude Code or Copilot, or talk to a program lead about scoping something specific to how your team actually uses it.

Choosing between tools? Read Cursor vs Claude Code vs Codex for teams.