AI Coding Assistants

Codex

OpenAI's coding agent with GPT-5.6 model choice, selectable reasoning, and parallel Ultra workflows.

Codex: OpenAI's coding agent with GPT-5.6 model choice, selectable reasoning, and parallel Ultra workflows. Pricing: From $8/mo + usage. Best for Repository-scale implementation, refactoring, debugging, testing, and review with tool use and Developers moving between the ChatGPT desktop app, web, terminal, IDE, and mobile Remote.

ChatGPT plan accessFrom $8/moChatGPT Business standard seatFrom $20/seat/moBusiness Codex seatTeam pricing

Free plan available

Overview

Bottom line

Codex is OpenAI's coding agent for planning, editing, running, and reviewing software work across the ChatGPT desktop app, web, terminal, IDE, and cloud-connected workflows. Its current model family is GPT-5.6: Sol for the hardest, most polished work, Terra for everyday tasks, and Luna for faster repeatable jobs. Developers can select reasoning effort, with Max extending one model's thinking and Ultra delegating parallel parts to subagents. It works best for developers and teams that want an agent to complete repository-scale implementations, refactors, debugging, tests, and reviews rather than only suggest the next line of code.

The main caveat is operational and financial complexity. Codex is joining the ChatGPT desktop app, while model choice, included usage, credits, and workspace controls vary by ChatGPT plan. ChatGPT Work shares Codex technology and the same agentic usage pool, but it is a broader work agent, not a replacement name for the coding product. Start with the included ChatGPT route and default reasoning setting, monitor the Usage dashboard, and move to higher effort or Ultra only when the task warrants it. API-key automation is a separate usage-based OpenAI API purchase path.

Pricing

How Codex pricing works

Access routes

Pricing may differ across app subscriptions, API usage, team workspaces, and enterprise contracts. See the pricing guide for current plan and meter details.

App subscriptionPrimary

ChatGPT plan access

From $8/mo

Codex is included across ChatGPT Free, Go, Plus, and Pro. Free and Go provide limited GPT-5.6 Terra access, while higher tiers expand model choice and usage; eligible Plus and Pro users can buy credits after included limits.

Team workspace

ChatGPT Business standard seat

From $20/seat/mo

Standard Business seats include ChatGPT and Codex with baseline usage plus optional shared workspace credits. Standard seats require at least two users and use fixed per-user billing.

Team workspace

Business Codex seat

Team pricing

Usage-based Codex-only access for a Business workspace, with no fixed seat fee or minimum. It requires workspace credits and does not include ChatGPT access.

Enterprise sales

Enterprise and Edu access

Custom

Organization access is contract-based, with workspace controls and shared credits governed by the applicable Enterprise or Edu agreement.

Direct API

OpenAI API

Usage-based API

API-key use is billed separately by model and token type. ChatGPT subscriptions and Codex credits do not include OpenAI API usage, and the Responses API multi-agent beta is a separate developer surface.

Capabilities

Input types

text, image, code, file, url

Output types

text, image, code, file

Supported platforms

Web, iOS, Android, Mac, Windows, Linux

Delivery modes

Web app, Mobile app, Desktop app, Chat app, CLI, IDE extension, SDK

Company context

Company

OpenAI

Founded

2015

Headquarters

San Francisco, California, USA

Official sources

Editorial guidance grounded in official product sources.

Codex official siteChecked July 12, 2026 UTC
Codex PricingChecked July 12, 2026 UTC
ChatGPT | ChatGPT LearnChecked July 12, 2026 UTC

FAQ

Common questions

Which GPT-5.6 models are available in Codex?

Free and Go users receive limited GPT-5.6 Terra access. Plus, Pro, Business, and Enterprise users can choose Sol, Terra, or Luna, subject to plan and workspace controls.

Are Max and Ultra the same Codex setting?

No. Max gives the selected model more time to reason about one task. Ultra uses subagents to work on separate parts in parallel and is available in Codex on Plus and higher plans.

Is Codex the same as ChatGPT Work?

No. ChatGPT Work is a broader agent for research, apps, files, and finished business materials. It uses Codex technology and shares the agentic usage pool, while Codex remains the developer-focused coding experience.

What happened to the standalone Codex desktop app?

Codex joined the ChatGPT desktop app on macOS and Windows. Existing users can update normally and keep their projects, settings, and workflows, with Codex remaining a dedicated view.

Does Codex have a published 400K context window?

Current Codex product documentation does not publish a Codex-specific 400K context limit. OpenAI API model context specifications should not be treated as Codex app limits.

How is Codex billed?

Codex starts with access and included usage through eligible ChatGPT plans. Extra usage may draw from credits where eligible, while API-key use is billed separately under OpenAI API pricing.

Recent updates

Changelog snapshot

Improvement

Event date August 24, 2026

Version 0.149.1

Codex CLI 0.149.1 shipped with the rust-v0.149.0...rust-v0.149.1 release changelog.

Official source ChatGPT & Codex changelog | ChatGPT Learn

VerifiedAugust 27, 2026

Feature

Event date August 20, 2026

Codex and ChatGPT updates added Apple Messages, Site co-editing, editable Site URLs, Computer History in Europe, and shared thread snapshots.

Official source ChatGPT & Codex changelog | ChatGPT Learn

VerifiedAugust 27, 2026

Decision rail

Keep the key facts, primary action, and section jumps visible while you evaluate the profile.

Visit Official Site

Last verified July 27, 2026

Pricing: From $8/mo + usage
Platforms: Web, iOS, Android, Mac, Windows, Linux
Access: Web app, Mobile app, Desktop app, Chat app, CLI, IDE extension, SDK
Best for: Repository-scale implementation, refactoring, debugging, testing, and review with tool use and Developers moving between the ChatGPT desktop app, web, terminal, IDE, and mobile Remote and Teams coordinating parallel worktrees, Ultra subagents, automations, and approval-aware workflows and OpenAI-centered teams that can keep ChatGPT access, Codex credits, and API spend separate
Category: AI Coding Assistants

On this page

Bottom line Pricing Capabilities FAQ Changelog Related tools

Pass this page along

Copy the link or send it to the channel where your team compares tools, pricing, and tradeoffs.

LinkedIn X Reddit Email

Navigation

Tool cluster

Open related pages for pricing, reviews, and adjacent coverage.

Review: Codex Review: Is GPT-5.6 Worth It for Developers?Codex pairs GPT-5.6 model choice with selectable reasoning and Ultra subagents, but its value depends on managing plan limits, credits, surfaces, and separate API billing.Pricing: Codex Pricing 2026: GPT-5.6 Plans, Credits, and API CostsCodex is included with ChatGPT Free, Go, Plus, Pro, Business, Edu, and Enterprise. ChatGPT Work shares Codex usage, while API-key workflows use separate token billing.Alternatives: Best Codex Alternatives in 2026Cursor, GitHub Copilot, and Windsurf are the clearest Codex alternatives if you want a more editor-first workflow, cheaper seat pricing, or a different balance between agent power and IDE convenience.

Head-to-head

Open direct comparison pages involving this tool.

VSCodex vs Claude CodeKeep the choice conditional. Claude Code is the better default for terminal-centered repository work, granular action permissions, checkpoints, and published Anthropic deployment routes. Codex is the better default for a unified workflow spanning local sessions, managed worktrees, cloud environments, subagents, and integrated review. Current official evidence does not justify turning this into an Opus 5 versus GPT-5.6 performance verdict.VSCodex vs CursorCursor is the better default for most developers because it combines an AI-native editor, broader model choice, and stronger day-to-day workflow tooling. Codex is stronger for multi-agent delegation, cross-tool reach, and direct OpenAI API usage.VSCodex vs GitHub CopilotCodex should be the default when the team is buying for deeper agentic coding work: repository understanding, task execution, command-running, tests, code review preparation, and reviewable diffs. GitHub Copilot should lead when the team values GitHub-native IDE assistance, pull request workflow, seat administration, and organization policy more than standalone agent depth.VSCodex vs Devin Desktop (Windsurf)Codex is the preferred default when the buyer wants an autonomous coding agent that can handle repo-level tasks, run commands, produce diffs, support cloud delegation, and plug into review workflows. Windsurf is the better switch when the real purchase is an AI-native IDE with autocomplete, Cascade, previews, local context awareness, model routing, and a tighter in-editor flow.

Same category

Nearby tools in the same lane.

ToolCursorAI code editor with agents, context-aware completion, Bugbot, and cloud workflows.ToolClaude CodeAnthropic’s coding agent for terminal, IDE, web, desktop, mobile, and SDK workflows ToolGitHub CopilotGitHub-native AI coding assistant for chat, code review, and agent workflows.ToolDevin Desktop (Windsurf)Windsurf is now Devin Desktop: an agentic IDE for local and cloud coding workflows.

Reference links

Official site Pricing page Documentation External changelog