ChatGPT Codex Goes Free: What Every User Gets in 2026

If you have ever wanted a coding agent that could write, debug, and review your code without charging you a subscription, OpenAI just made that possible. On May 14, 2026, the company quietly rolled out Codex Mobile to iOS and Android and simultaneously extended full Codex access to every ChatGPT plan -- including the free tier that over 400 million people already use. This is not a limited preview you need to join a waitlist for. It is live, accessible today, and it works on web, CLI, IDE extension, and mobile. Whether you are a student learning Python, a solo developer maintaining a side project, or someone who just wants to automate a repetitive script, Codex is now a free tool in your arsenal.

This article breaks down exactly what Codex does, what the free plan actually gives you, how limits compare across tiers, and what you need to do to get started right now.

What Is ChatGPT Codex and What Does It Actually Do?

Codex is OpenAI's dedicated AI coding agent, built for software engineering tasks that go well beyond simple autocomplete. It was first released as a command-line tool (Codex CLI) in April 2025, and has since expanded into a full product suite with a web app, IDE extensions for VS Code and JetBrains, and as of May 2026, a mobile app.

Unlike the coding help you get inside a standard ChatGPT conversation -- where you paste in a function and ask for changes -- Codex operates as an autonomous agent. It can read your entire repository context, understand how different files relate to each other, and take multi-step actions to complete a task. Concretely, here is what you can hand it:

• Write new code -- Describe a feature or function in plain language; Codex generates working code in your chosen language and file structure.
• Review existing code -- Codex analyzes your codebase to identify potential bugs, logic errors, and edge cases you may have missed.
• Debug and fix problems -- When a test fails or something breaks, Codex traces failures, diagnoses root causes, and suggests targeted fixes rather than asking you to start over.
• Generate and run tests -- Codex can write unit tests, integration tests, and run them against your code in a sandboxed environment (cloud tier and above).
• Refactor and optimize -- Hand it a messy legacy function and ask for a cleaner rewrite; Codex preserves intent while improving structure and performance.

The agent runs locally for most free and Go tier tasks -- meaning your code and prompts are processed without needing a persistent cloud environment. Paid tiers unlock cloud execution, which allows longer-running tasks, automatic GitHub code review via pull requests, and Slack integration for team notifications.

Is Codex Really Free on ChatGPT's Free Plan?

Yes, and OpenAI's own documentation confirms it. According to the OpenAI Codex documentation, "Codex is included across Free, Go, Plus, Pro, Business, Edu, and Enterprise plans." The free plan is explicitly listed first, with no asterisk or waitlist condition attached.

To access it, you sign in to your existing ChatGPT account (no new signup required) and connect to one of the four Codex clients: the web app at chatgpt.com/codex, the CLI, an IDE extension, or the mobile app. The web and mobile clients require connecting your GitHub account to enable full context loading -- which is optional but recommended if you want Codex to actually see your project files.

What "free" means in practice is a limited but real allocation of coding tasks. The free tier is positioned around quick, exploratory tasks -- think writing a helper function, fixing a failing test, or converting a class-based React component to functional style. It is not designed for all-day autonomous development sessions, but it is more than enough to get meaningful work done without opening your wallet.

One important note: Codex usage on the free plan counts against your shared agentic usage limit, which is also consumed by ChatGPT for Excel and any Workspace Agents you run. If you are a heavy user of those tools, you may find your Codex allocation depleted faster than expected.

What Are the Usage Limits for Each ChatGPT Plan?

According to OpenAI's developer pricing page, Codex usage is governed by rolling 5-hour windows rather than monthly caps. This means limits refresh throughout the day as you use them, which is more practical for developers who work in bursts.

The exact number of tasks you can run depends heavily on the model selected and the complexity of the coding task. A small function with minimal context may use only a fraction of your limit, while a complex refactor across a 50-file codebase could consume several times as much per task. Here is how the tiers break down:

It is worth noting that OpenAI has made an efficiency argument for the newer GPT-5.5 model: it uses significantly fewer tokens to achieve results comparable to GPT-5.4, which is why the 5.5 limits look lower on paper but deliver comparable output. For most free tier use cases, the 15-80 window on GPT-5.5 translates to a practical coding session without feeling artificially truncated.

What Changed on May 14, 2026?

The May 14 announcement was a two-part expansion. First, OpenAI formally extended Codex access to the free and Go tiers -- previously, Codex had been available only to Plus and above. Second, the company launched Codex Mobile as a preview on iOS and Android, available on every plan including free.

The mobile launch is particularly significant for how developers actually work in 2026. With Codex on iOS and Android, you can review a failing CI job from your phone during a commute, assign a debugging task to the agent, and come back to a fixed PR by the time you reach your desk. The mobile app connects to the same cloud infrastructure as the desktop clients, so tasks assigned on mobile continue running in the background even after you close the app -- assuming you are on a plan that supports cloud execution.

According to reporting from FelloAI's ChatGPT pricing guide, the Codex Mobile rollout was part of a broader May 2026 product push that also included ChatGPT Personal Finance for Pro subscribers and the expansion of OpenAI's Ads Manager for the free tier. The pattern is clear: OpenAI is expanding the free tier's feature set while monetizing through ads and higher-tier subscriptions rather than locking basic functionality behind a paywall.

How Does Free Codex Compare to the Paid Alternatives?

The honest answer is that free Codex is a meaningful tool with real limitations, not a crippled demo. The key dividing line between free and Plus is not model quality -- both tiers access GPT-5.5 for local tasks -- it is the absence of cloud features on the free tier. Cloud execution unlocks three things that matter for professional workflows:

• Automatic GitHub code review -- Codex can act as a reviewer on pull requests, flagging potential issues before human review begins. This is unavailable on free.
• Persistent cloud tasks -- On Plus and above, you can assign a long-running task (like refactoring an entire module) and let it run in the background. Free users can only run local tasks that complete within a single session.
• Slack integration -- Teams using Slack can receive Codex output and notifications directly in channels, useful for shared development environments. This requires Plus or higher.

For comparison, GitHub Copilot's free tier (also expanded in 2026) offers 2,000 code completions and 50 chat messages per month, while free Codex offers task-based usage measured in a rolling 5-hour window. The two tools solve slightly different problems -- Copilot is optimized for inline completions as you type, while Codex is designed for discrete, multi-step agentic tasks. Many developers use both, treating them as complementary rather than competitive.

If you want to track how all the major AI coding tools compare on free tier limits, the Free Tier Tracker keeps a current comparison of Codex, Copilot, Cursor, and other AI coding assistants. And for a deeper comparison of what you actually get per dollar, see our AI Pricing Comparison.

How Do You Get Started With Codex on the Free Plan?

Getting started takes about five minutes if you already have a ChatGPT account. Here is the fastest path:

1. Sign in at chatgpt.com -- Use your existing free account. No upgrade required.
2. Navigate to chatgpt.com/codex -- This takes you directly to the Codex web client, which is the easiest starting point for new users.
3. Connect GitHub (optional but recommended) -- Codex can work without a GitHub connection, but connecting one lets it read your repositories and give context-aware answers rather than generic ones.
4. Describe your first task -- Be specific: instead of "help me with my code," try "find any null pointer risks in this function and suggest fixes." Codex performs best with clear, bounded task descriptions.
5. Install the CLI for terminal work -- If you prefer working in your terminal, the Codex CLI (`npm install -g @openai/codex`) runs locally and authenticates against your ChatGPT plan the same way.

One tip for getting the most out of the free tier's limited window: use GPT-5.5 for most tasks since it is the most efficient model, consuming fewer tokens per result than GPT-5.4. Save the mini model (GPT-5.4-mini) for simple, repetitive tasks where you do not need frontier-level reasoning -- it has a higher task limit and costs less against your allocation.

You can also explore free AI coding tools side-by-side on our Free AI Tools directory, which covers Codex alongside other agentic coding assistants that do not require a subscription.