What does ChatGPT Codex cost in 2026? Pricing and access options explained
Quick Answer: ChatGPT Codex is bundled into ChatGPT plans rather than sold standalone. As of July 2026 the lineup is Free ($0, limited Codex), Go ($8/mo), Plus ($20/mo), Pro ($100/mo for 5x usage or $200/mo for 20x), Business ($25/user/mo, or $20 annual, minimum 2 users), and Enterprise (custom) — the old Team $30 plan was replaced by Business on April 2, 2026. Codex now runs on the GPT-5 series (GPT-5.5, GPT-5.4, GPT-5.4 mini) and is billed against credit/token usage, so real-world cost runs roughly $100-$200 per active developer per month.
Current Billing (verified July 2026)
Codex is included across the Free, Go, Plus, Pro, Business, and Enterprise plans rather than sold as a standalone product. Two things changed in 2026 that older guides get wrong:
- The plan lineup changed. A new Go tier ($8/month) sits below Plus, the Team $30/user plan was replaced by Business at $25/user/month (about $20/user/month billed annually, minimum 2 users) on April 2, 2026, and Pro now splits into two usage tiers: $100/month (5x Plus usage) and $200/month (20x Plus usage).
- The models changed. Codex now runs on the GPT-5 series — GPT-5.5, GPT-5.4, and GPT-5.4 mini — rather than the earlier codex-1 (o3-family) model.
Since April 2, 2026 Codex usage on Plus, Pro, Business, and Enterprise is billed against credit/token consumption rather than per message. Per 1M tokens the credit rates are GPT-5.5 = 125 in / 750 out, GPT-5.4 = 62.50 / 375, and GPT-5.4 mini = 18.75 / 113. Typical real-world usage runs about $100 to $200 per active developer per month, varying with model, parallel instances, and fast mode. (Source: OpenAI Codex pricing, accessed July 2026.)
Plan pricing (as of July 2026)
| Plan | Price | What you get |
|---|---|---|
| Free | $0 | Codex for quick coding tasks with limited usage; buy extra credits at the cap |
| Go | $8/month | Codex included for lightweight tasks; higher limits than Free |
| Plus | $20/month | Codex across web, CLI, IDE extension, and iOS; ~15-80 GPT-5.5 messages per 5-hour window |
| Pro (5x) | $100/month | 5x Plus rate limits; ~75-400 GPT-5.5 messages per 5h |
| Pro (20x) | $200/month | 20x Plus rate limits; the highest individual usage |
| Business | $25/user/month ($20 annual) | SAML SSO, MFA, no default training on your data; Codex via workspace credits; min 2 users |
| Enterprise | Custom | SCIM, EKM, RBAC, audit logs, data residency; Codex via workspace credits |
The Codex CLI remains a free, open-source terminal tool that connects with your own API key; using it that way bills token usage directly through the OpenAI API rather than a subscription.
Access options and cost control
- Subscription (Plus/Pro/Business): Codex draws on a shared usage allowance also used by other agentic ChatGPT features. Plus and Pro users can buy additional credits once they hit the cap.
- API key (unattended automation): billed strictly per token at standard OpenAI API rates, with no ceiling — set a hard budget cap.
- Choosing a tier: individual developers are well served by Plus at $20; heavy agentic users move to Pro (5x at $100 before 20x at $200); small teams that want SSO and no-training guarantees use Business at $25/user.
How it compares
At the entry level Plus ($20/month) is priced with GitHub Copilot Pro ($10/month), Claude Pro ($20/month), and Cursor Pro ($20/month). The distinction is interface and autonomy: Codex runs agentic, sandboxed tasks (including cloud tasks that open pull requests), whereas Copilot and Cursor are IDE-embedded. For unattended or high-volume generation, the API path is usually cheaper per unit of output than any subscription, at the cost of building your own tooling.
Editor's Note: Across ShadowGen engagements using Codex in 2026, the April-2026 switch from per-message to token-based credits changed how we budget it. On one client's migration work, two engineers running parallel Codex sessions against a large codebase exhausted the shared five-hour usage window by early afternoon, because cloud tasks and the local CLI both draw from the same allowance. Re-checked July 2026: the new Go ($8) and Business ($25/user) tiers reshuffled the low end — for a small team that wants SSO and a no-training guarantee, Business is now the sensible floor rather than stacking individual Plus seats. For unattended automation we still use an API key with a hard budget cap, since that path bills real dollars per token with no ceiling. — Rafal Fila, ShadowGen
Related Questions
Related Tools
ChatGPT Codex
OpenAI's cloud-based autonomous coding agent integrated into ChatGPT
AI Coding & Development ToolsClaude Code
Anthropic's agentic CLI tool for AI-assisted coding and automation development
AI Coding & Development ToolsBardeen
AI-powered browser automation via Chrome extension
Workflow AutomationAider
Open-source command-line AI pair programmer that edits Git repositories with multi-file context and automatic commits.
AI Coding & Development ToolsRelated Rankings
Best AI App Builders in 2026
AI app builders are a 2024-2026 category of products that turn natural-language prompts into deployable web applications. The category emerged from the convergence of frontier LLM capability (Claude, GPT-4o, Gemini) and improved tooling for code generation, in-browser runtimes (WebContainers), and managed application hosting. This ranking evaluates 7 platforms on output quality, deployment options, pricing, stack flexibility, and the underlying AI model quality. The ranked products span dedicated AI app builders (Lovable, Bolt.new, v0, Magic Loops), in-browser agentic IDEs (Cursor, Replit Agent), and autonomous coding agents (Devin). Scores reflect hands-on evaluation of each platform's ability to generate, run, and deploy a real web application from a prompt as of May 2026.
Best AI Coding Tools and Developer Assistants 2026
AI coding tools have become essential for professional developers in 2026, with the category spanning full AI-native editors, IDE plugins, terminal-based assistants, and code generation platforms. This ranking evaluates the leading AI coding tools based on code suggestion quality, IDE integration depth, programming language support, pricing value, and AI model quality. The evaluation focuses on tools that directly assist developers in writing, refactoring, and understanding code. General-purpose AI chatbots that can discuss code but do not integrate into development environments are excluded.
Dive Deeper
Claude Code vs ChatGPT Codex vs Cursor 2026: Three-Way Comparison
Claude Code (terminal CLI), ChatGPT Codex (cloud sandbox), and Cursor (VS Code fork) take three different approaches to AI-assisted coding. This three-way comparison covers pricing, autonomy, form factor, context handling, and agentic capabilities as of May 2026 to help engineers pick the right tool for each task class.
Lovable vs Bolt.new 2026: AI App Builders Compared
Lovable (Stockholm, 2023) ships React + Supabase apps with GitHub export from $25/month per-message. Bolt.new (StackBlitz, 2024) generates apps in-browser via WebContainers from $20/month per-token. This 2026 comparison covers stack, deployment, pricing, and which builder fits which use case.
Aider vs Cline 2026: Open-Source AI Coding Compared
Aider and Cline are two open-source AI coding tools that share a bring-your-own-key philosophy but ship in different form factors. Aider is a Python terminal CLI that pairs with developers via diffs and auto-commits; Cline is a VS Code extension that runs an autonomous coding agent. As of April 2026 both are Apache 2.0 licensed, free to install, and bill the developer's model API directly.