Modelpeers
OpenAI Codex charactervsClaude Code character

OpenAI Codex vs Claude Code: Which coding agent should you use?

Facts checked Oct 22, 2026. Every number links to its source.

Summary

Derived from sourced facts, not from builder opinions
  • On the official Terminal-Bench 4.0 board, which tests each agent with a chosen model, Claude Code with Opus 5.5 leads at 64.85%. Codex scored 58.18% with GPT-6 Astra or GPT-6.1 Sol, all at max effort.
  • Both start at $20 a month: Claude Pro and ChatGPT Plus. Codex also has Free and $8 Go tiers. Claude Code is not in Claude's free plan.
  • Codex publishes usage ranges per model (Plus: about 15-160 messages per 5 hours with GPT-6.1 Sol), shared with ChatGPT Work. Claude Code publishes rolling 5-hour and weekly windows without per-plan numbers.
  • Claude Code is built for Claude models. Other makers' models need an API that accepts Anthropic's format, set up and billed separately. Codex runs OpenAI models only.

Choose OpenAI Codex if

  • You want published usage numbers per model.
  • You want a free or $8 entry tier.
  • Your team already pays for ChatGPT.
  • You want OpenAI's GPT-6 models. Claude Code's built-in models are Claude models only.

Choose Claude Code if

  • You want the top Terminal-Bench 4.0 result (Claude Code with Opus 5.5).
  • You use Claude through Bedrock, Google Cloud or Foundry.
OpenAI Codex
Claude Code

Follow both to get an email when either one changes its price, limits or behavior.

Plans

OpenAI Codex plans

Free
Free

GPT-6 Luna in the desktop app, rolling out

Go
$8/month

GPT-6 Luna in the desktop app, rolling out

Plus
$20/month

Web, CLI, IDE and iOS · GPT-6.1 Sol and GPT-6 Luna

Pro
$100/month

Plans at $100, $200 or $500 · no five-hour limit

Business
$25/month

Per user monthly · $20 billed annually

Claude Code plans

Pro
$20/month

Claude Code included · billed $17/month annually

Max
$100/month

From $100: 5x or 20x Pro usage

Team Standard seat
$25/month

Per seat, monthly · $20 billed annually

Team Premium seat
$125/month

5x a standard seat · $100 billed annually

Enterprise
$20/month

Per seat billed annually, 20-seat minimum, plus usage at API rates

Subscription agents are priced per seat with usage limits, so this compares plans rather than dividing a subscription by a guessed number of tasks.

Side by side

OpenAI CodexClaude Code
MakerOpenAI
Anthropic
PlansFree $0 and Go $8 (GPT-6 Luna in desktop app, rolling out) · Plus $20 · Pro $100, $200 or $500 · Business $20/user annual or $25 monthly · Enterprise and Edu by sales · API key pay-as-you-go
Not in Free. Pro $20/month ($17 annual) · Max from $100 (5x or 20x Pro usage) · Team Standard $25/seat monthly ($20 annual), Premium $125 ($100 annual) · Enterprise $20/seat plus usage at API rates
Usage limitsPlus and Standard Business local messages per 5 hours: GPT-6 Astra 5-45, GPT-6.1 Sol 15-160, GPT-6 Sol 15-150, GPT-6 Luna 350-3,000 · Pro has no 5-hour limit · weekly limits may apply
Per-seat 5-hour and weekly windows · weekly Claude Code limits 25% above pre-May 2026 levels since 2026-09-14 · no published per-plan numbers
Fast modesFast mode uses included limits at 2.5x · Ultrafast for Astra and 6.1 Sol at 8x
Not published
Recent model retirementGPT-5.5 retired from Codex with ChatGPT sign-in on 2026-10-14
Not published
PlatformsWeb, CLI, IDE extension, iOS, desktop app · API key mode for CLI, SDK and IDE without cloud features
Terminal CLI, VS Code, JetBrains, desktop app (macOS, Windows, Linux beta), web, iOS and Android app
Quality tracking (MarginLab)Codex with GPT-6 Sol/high: 54% latest day, 59% 7-day, 79% 30-day · 1-day and 7-day drops flagged but detection paused (last updated 2026-09-29)
Not published
ModelsNot publishedBuilt for Claude models: aliases opus (Opus 5.5), sonnet (Sonnet 5.5), haiku (Haiku 5.5), fable (Fable 5.1, may bill usage credits) · third-party providers Bedrock, Google Cloud, Foundry
Use with other toolsNot publishedSubscription OAuth may not be used to route requests for third-party products
Terminal-Bench 4.0 (official board)58.18%with GPT-6 Astra (max) and with GPT-6.1 Sol (max), tied at rank 364.85%with Opus 5.5 at max effort, rank 1 (posted 2026-10-06)

Scores appear side by side only when both were published for the same test. Settings differ between sources, so small gaps can reverse under another harness.

Updates

OpenAI Codex

12 recorded incidents in the last 90 days.

Claude Code

27 recorded incidents in the last 90 days.

Switching checklist

  1. Install Claude Code with curl -fsSL https://claude.ai/install.sh | bash on macOS or Linux, or irm https://claude.ai/install.ps1 | iex in Windows PowerShell, then run claude in your project.
  2. Sign in with a Pro, Max, Team or Enterprise plan, or a Console account. The free Claude plan does not include Claude Code.
  3. Keep your AGENTS.md. Claude Code reads it as the project instructions when the repository has no CLAUDE.md (version 2.1.277 or later).
  4. Run /init to generate a CLAUDE.md from your codebase, or add one that imports AGENTS.md if you want Claude-only instructions on top.
  5. Move MCP servers out of the [mcp_servers] tables in ~/.codex/config.toml.
  6. Recreate your MCP servers with claude mcp add, or claude mcp add-json with the object inside mcpServers. Use --scope project to write a shared .mcp.json, and add "type": "http" to any url entry.
  7. Review permissions: Claude Code reads files without asking and asks before it edits files or runs most commands, close to Codex's default workspace-write with on-request approvals.
  8. Pick a model with /model: opus (Opus 5.5) is the default on paid plans, with sonnet, haiku and fable also available.
  9. Run one real task in your own repository with both tools and compare the diffs before moving a team.

FAQ

Is Claude Code better than Codex?

On the official Terminal-Bench 4.0 board, Claude Code with Opus 5.5 scored 64.85% and Codex scored 58.18% with GPT-6 Astra or GPT-6.1 Sol, all at max effort. No other shared test covers both agents.

How much do Claude Code and Codex cost?

Both start at $20 a month (Claude Pro, ChatGPT Plus). Codex also has Free and $8 Go tiers and Pro plans at $100, $200 or $500.

What are the Codex usage limits?

On Plus, about 15-160 local messages per 5 hours with GPT-6.1 Sol and 5-45 with GPT-6 Astra. ChatGPT Work shares the same allowance. Pro has no five-hour limit.

Can Claude Code use GPT models?

Not as a built-in choice. Claude Code's model settings list Claude models through the opus, sonnet, haiku and fable aliases. Other makers' models need an API that accepts Anthropic's format, set up and billed separately.

Can Claude Code read my AGENTS.md?

Yes. Claude Code reads AGENTS.md as the project instructions when the repository has no CLAUDE.md, from version 2.1.277.

Is Claude Code free?

No. It starts with Claude Pro at $20 a month, or $17 billed annually.

Where can I use Claude Code?

In the terminal, VS Code, JetBrains, the desktop app, the web and the iOS and Android app.

Related comparisons

Cursor vs Claude CodeOpenClaw vs Claude CodeOpenAI Codex vs CursorClaude Code vs GitHub CopilotOpenAI Codex vs GitHub CopilotClaude Code vs ChatGPTGoogle Antigravity vs Claude CodeGemini CLI vs Claude CodeOpenAI Codex pageClaude Code page

Sources and method

Prices, specs and limits come from each maker's own pages. Benchmark scores come from independent benchmark publishers and the makers' announcements, with the setting each was measured in. Changes come from changelogs, deprecation pages and status pages. The summary is written from these facts only.