OpenAI Codex is OpenAI's coding agent, available as an open-source CLI, an IDE extension and a cloud agent that works on tasks in parallel sandboxes. It is included with ChatGPT plans — from limited use on Free to Plus ($20/month) and Pro ($100–$500/month).
What is OpenAI Codex?
Codex can run locally (CLI or IDE extension, editing files on your machine) or in the cloud, where each task gets its own sandboxed environment and returns a diff or pull request. Cloud features include automatic GitHub code review and a Slack integration.
Since April 2026 Codex usage is metered in credits aligned to token usage; plans include an allowance and you can buy more credits. You can also run the CLI with a plain OpenAI API key and pay per token.
OpenAI Codex pricing
Plans as publicly listed on the official pricing page, verified 2026-10-10.
| Plan | Price | Notes |
|---|---|---|
| ChatGPT Free / Go | Included (limited) | Codex is included in all ChatGPT plans |
| Plus | $20/mo | Codex web, CLI, IDE; code review & Slack integrations |
| Pro | $100, $200 or $500/mo | Much higher usage; Ultrafast mode on $500 |
| Business / Enterprise | Per seat | Check official site |
| API key | Pay per token | CLI/IDE only, no cloud features |
Source: https://chatgpt.com/codex/pricing/. Prices in USD, excluding tax.
Supported models
OpenAI GPT models (current GPT-6 family in Codex). Bring-your-own-key: supported.
OpenAI
Features at a glance
- ✓ Agent mode
- — Autocomplete
- ✓ Cloud / background agents
- ✓ PR code review
- ✓ MCP support
- — Self-host option
Interfaces: Terminal (Codex CLI) · IDE extension · Web / ChatGPT · Desktop app · iOS.
Strengths and weaknesses
＋ Strengths
- Included in ChatGPT subscriptions many developers already have
- Parallel cloud tasks that return PRs
- Open-source CLI
- Strong agentic benchmark results with current OpenAI models
－ Weaknesses
- OpenAI models only
- Credit-based usage rules are complex and change often
- Cloud features unavailable with API-key-only use
Who is OpenAI Codex best for?
ChatGPT subscribers who want to delegate well-scoped tasks to background agents and review PRs.
OpenAI Codex benchmark results
Independent agent-level results we track. Scores depend on the model and harness — see how to read benchmarks.
| Benchmark | Score | Setup | Date |
|---|---|---|---|
| Terminal-Bench 4.0 | 58.2% | with GPT-6 Astra (max effort) | Oct 2026 · source |
| Terminal-Bench 2.1 | 89.5% | agent-level, Artificial Analysis harness | Aug 2026 · source |
OpenAI Codex alternatives
Developers who are comfortable in the terminal and want an agent that can own multi-step tasks end to end.
from $20/mo · Compare →
Developers who want to offload chores (dependency bumps, tests, bug fixes) to a background agent for free.
Free tier · paid: see site · Compare →
Teams with a backlog of well-defined tickets they want to delegate.
Free tier · from $20/mo · Compare →
Teams already on GitHub who want AI across editors and the PR workflow at a low price.
Free tier · from $10/mo · Compare →
Head-to-head: Claude Code vs Codex (in-depth) · OpenAI Codex vs Claude Code · OpenAI Codex vs Jules · OpenAI Codex vs Devin · OpenAI Codex vs GitHub Copilot