ChatGPT Codex

by OpenAI

Cloud freemium API Available
Developer-Friendly

OpenAI's cloud-based autonomous coding agent integrated into ChatGPT ChatGPT Codex is OpenAI's cloud-based software-engineering agent, launched in May 2025 as a feature inside ChatGPT for Plus, Pro, Team, Enterprise, and Edu subscribers. The product is OpenAI's second use of the Codex name; the original 2021 Codex API (which powered the first version of GitHub Copilot) was deprecated in March 2023 and is unrelated to the May 2025 product.

Performance Scores

7.9

1 ranking evaluated

Score range: 7.9 – 7.9

Key Facts

technical

technical facts about ChatGPT Codex
AttributeValueAs ofSource
Underlying modelcodex-1, an OpenAI o3-family model specialised for software engineering via reinforcement learningMay 2026OpenAI Codex launch announcement
Codex models + credit ratesCodex 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 model. Usage is credit/token-metered: per 1M tokens GPT-5.5 = 125 credits in / 750 out; GPT-5.4 = 62.50 / 375; GPT-5.4 mini = 18.75 / 113.Jul 2026OpenAI Codex pricing

product

product facts about ChatGPT Codex
AttributeValueAs ofSource
Launch dateLaunched May 2025 inside ChatGPT for Plus, Pro, Team, Enterprise, and Edu subscribersMay 2026OpenAI Codex launch announcement

pricing

pricing facts about ChatGPT Codex
AttributeValueAs ofSource
Codex plan pricing (July 2026)Codex is included across ChatGPT plans (verified July 2026): Free $0, Go $8/mo, Plus $20/mo, Pro $100/mo (5x usage) or $200/mo (20x usage), Business $25/user/mo ($20 annual, min 2 users), Enterprise custom. The Team $30 plan was replaced by Business on April 2, 2026.Jul 2026OpenAI Codex pricing
Usage limit (June 2026)Codex shares an "agentic usage limit" with ChatGPT for Excel and Workspace Agents; Plus and Pro users can buy extra credits at the capJun 2026OpenAI Help Center
Typical monthly cost (June 2026)Typical real-world Codex cost is about $100 to $200 per developer per month, varying by model, parallel instances, and fast modeJun 2026OpenAI Codex Pricing
Billing model (June 2026)Since April 2, 2026, Codex is billed on API token usage (credits per million input/cached/output tokens) on Plus, Pro, Business, and new Enterprise, replacing per-message pricingJun 2026OpenAI Codex Pricing
Access (May 2026)Included with ChatGPT Plus ($20/mo), Pro ($200/mo), Team (from $25/seat/mo annual), EnterpriseMay 2026ChatGPT pricing page

General

General facts about ChatGPT Codex
AttributeValueAs ofSource
ExecutionCloud tasks run in isolated cloud environments; Codex also runs locally via CLI and IDE extension. The ChatGPT iOS/Android app provides remote monitoring/control of Codex tasks (approvals, model changes, outputs) but does not execute tasks on the phone (as of July 2026)Jul 2026OpenAI Codex documentation
ModelGPT-5.6 model family (default gpt-5.6-sol, with Terra and Luna tiers); original codex-1 (o3-based) retired (as of July 2026)Jul 2026OpenAI Codex models documentation
Parallel ExecutionMultiple tasks execute simultaneously in separate cloud containersJul 2026OpenAI Codex cloud documentation
GitHubCreates pull requests, commits, and branches directly on GitHubJul 2026OpenAI Codex cloud documentation
PricingIncluded with all ChatGPT plans: Free (limited), Go $8/month, Plus $20/month, Pro $100/month (5x) or $200/month (20x), Business $25/user/month ($20 annual), Enterprise custom (as of July 2026)Jul 2026OpenAI Codex pricing documentation
Release DateMay 2025 (research preview)Jul 2026OpenAI — Introducing Codex
SandboxEach task runs in a pre-configured cloud environment with project dependenciesJul 2026OpenAI Codex cloud documentation
OperationsCode writing, bug fixing, test generation, PR creation, refactoringJul 2026OpenAI Codex cloud documentation
AsyncTasks run asynchronously — results available when complete, no waiting requiredJul 2026OpenAI Codex cloud documentation
MCP SupportSupports MCP — connects to local STDIO and remote HTTP MCP servers via Codex CLI, IDE extension, and ChatGPT desktop app, configured in ~/.codex/config.toml (as of July 2026)Jul 2026OpenAI Codex MCP documentation

Strengths

  • Frequent model upgrades
  • Multiple execution surfaces

Limitations

  • Capability varies with model version
  • Less established agent UX than rivals

Based on evaluations in 1 ranking: Best AI Coding Tools and Developer Assistants 2026

Pricing Plans

Free

Free

Codex included with limited usage

  • Codex for quick coding tasks
  • Buy extra credits at the cap
  • !Limited usage
Get started →
As of Jul 2026 · Source

Go

$8.00/mo

Codex included; lightweight tasks

  • Codex included
  • Higher limits than Free
Get started →
As of Jul 2026 · Source
Most Popular

Plus

$20.00/mo

Per user/month

  • Codex across web, CLI, IDE, iOS
  • GPT-5 series Codex models
  • !~15-80 GPT-5.5 messages / 5h window
Get started →
As of Jul 2026 · Source

Pro (5x)

$100.00/mo

Per user/month; 5x Plus usage

  • 5x Plus rate limits
  • !~75-400 GPT-5.5 messages / 5h
Get started →
As of Jul 2026 · Source

Pro (20x)

$200.00/mo

Per user/month; 20x Plus usage

  • 20x Plus rate limits
  • Highest individual usage
  • !~300-1600 GPT-5.5 messages / 5h
Get started →
As of Jul 2026 · Source

Business

$25.00/mo

(or $20.00/yr)

$25/user monthly, $20/user annual; min 2 users

  • SAML SSO, MFA
  • No default training on your data
  • Codex via workspace credits
  • !2+ users
Get started →
As of Jul 2026 · Source

Enterprise

Contact us

Contact sales

  • SCIM, EKM, RBAC, audit logs
  • Data residency
  • Codex via workspace credits
Get started →
As of Jul 2026 · Source

About ChatGPT Codex

ChatGPT Codex is OpenAI's cloud-based software-engineering agent, launched in May 2025 as a feature inside ChatGPT for Plus, Pro, Team, Enterprise, and Edu subscribers. The product is OpenAI's second use of the Codex name; the original 2021 Codex API (which powered the first version of GitHub Copilot) was deprecated in March 2023 and is unrelated to the May 2025 product. The 2025 Codex agent is anchored on codex-1, an OpenAI o3-family model specialised for coding tasks through additional reinforcement-learning training on real-world software-engineering trajectories.

The product runs in cloud-hosted, isolated containers preloaded with the user's repository code. The agent can read and edit files, run terminal commands, and execute tests inside the sandbox; network access during agent execution is configurable per workspace. Each task spins up its own ephemeral container, and the agent reports progress through a streaming interface inside ChatGPT. Output options include direct commits and pull-request authoring against connected GitHub repositories. The May 2025 launch positioned Codex as a "delegate-and-review" workflow rather than a real-time pair-programmer, which is the niche occupied by GitHub Copilot and Cursor.

Codex is included at no additional cost in ChatGPT Plus ($20/month), Pro ($200/month), Team (from $25/seat/month annual), Enterprise, and Edu plans during the rollout phase OpenAI announced in May 2025. The company has stated that usage limits and longer-term pricing structure may evolve as the product moves out of initial release; through 2026 the practical bottleneck has been compute-time fair-use limits inside ChatGPT Plus rather than feature gates. A standalone Codex CLI (open-source, MIT) provides terminal access to the same model and tooling pattern for users who prefer command-line workflows.

OpenAI has positioned Codex against three categories: traditional code-completion tools (Copilot, Cursor, Windsurf, Codeium), terminal-native AI coding tools (Claude Code, Aider, Cline), and other cloud-hosted coding agents (Devin, Replit Agent, Codex-CLI itself). The differentiator OpenAI emphasises is the isolated-sandbox execution model: agents complete entire tasks, including running tests, before returning a PR for review, rather than streaming suggestions into the developer's editor. The practical question for buyers in 2026 is whether their workflow benefits more from real-time inline assistance or delegated background execution; in most ShadowGen engagements the answer has been "both, used in different contexts."

flowchart LR
  A[Developer prompt in ChatGPT] --> B[Codex Agent]
  B --> C[(Ephemeral Sandbox: Repo Clone)]
  C --> D[Read / Edit Files]
  C --> E[Run Tests in Sandbox]
  C --> F[Execute Shell Commands]
  D --> G{Progress Stream}
  E --> G
  F --> G
  G --> H[Pull Request on GitHub]
  G --> I[Review in ChatGPT]

Editor's Note: Across 8 ShadowGen client teams who trialled Codex against Claude Code and Cursor in late 2025 and early 2026, the practical pattern that emerged is splitting work across tools rather than picking one. Codex performs best on well-scoped, test-covered tickets (median ShadowGen acceptance rate: 64% on first PR, n=147 tickets) and noticeably worse on tickets requiring multi-file architectural reasoning across legacy codebases. Cursor remains the better real-time editor companion; Claude Code remains the better terminal-native agent. We have not yet seen a single coding tool replace the other two in production engineering teams. — Rafal Fila, ShadowGen

Integrations (4)

Cloud Sandbox native
Docker native
GitHub native
VS Code api

Written & reviewed by Rafal Fila · Last updated: · Last verified:

Other AI Coding & Development Tools Tools

See How It Ranks

Questions About ChatGPT Codex

How much do AI coding assistants cost in 2026?

As of June 2026, mainstream AI coding assistants cluster in two cost shapes. Per-seat subscriptions with included AI usage: GitHub Copilot Pro $10/month (Business $19/seat), Cursor Pro $20/month, and Claude Code and ChatGPT Codex bundled into Claude ($20+) and ChatGPT ($20+) subscriptions. Free, bring-your-own-model tools where you only pay API spend: Aider and Cline ($0 for the tool, roughly $5-30/day in model cost for active use). Replit Agent is credit-metered from $25/month. The 2026 catch is that most paid tiers moved to usage metering, so the sticker price is a floor, not a ceiling.

Claude Code vs Codex vs Cursor for autonomous coding in 2026: which fits best?

For terminal-first developers and shell-heavy refactors, Claude Code (Anthropic, $20-200/month) is the strongest fit. For background, async, end-to-end task completion with PRs, ChatGPT Codex ($20-200/month bundled with ChatGPT) wins on autonomy. For real-time IDE pair programming inside a VS Code fork, Cursor ($20-40/user/month) is the most ergonomic. Most 2026 teams use two or three of them in parallel, assigned to different task classes.

What are the best AI coding tools in 2026?

The leading AI coding tools in 2026 are [GitHub Copilot](/tools/github-copilot/) (default for GitHub-centric teams, $19/user/month Business), [Cursor](/tools/cursor/) (AI-native VS Code fork with strong codebase context), and [Claude Code](/tools/claude-code/) (Anthropic terminal coding agent for repository-wide refactors).

What are the best ChatGPT Codex alternatives in 2026?

As of April 2026, the leading ChatGPT Codex alternatives are Claude Code (terminal and IDE agent from Anthropic), Cursor (AI-first IDE with Composer), GitHub Copilot (Microsoft baseline with agent mode), Aider (open-source CLI for git-based edits), and Cline (open-source VS Code extension that pairs with any model provider).

Learn More