Codex vs Claude Code: prices, limits, models and benchmarks

Codex vs Claude Code at a glance

When you compare Codex vs Claude Code, full access to both starts at $20 a month with ChatGPT Plus or Claude Pro, though ChatGPT Free and Go users also get GPT-6 Luna in Codex's desktop app. The real differences come down to the models each tool runs, how OpenAI and Anthropic publish usage limits, and the CLI license.

Prices and plans

On the Codex pricing ladder, ChatGPT Free and Go ($8 a month) get GPT-6 Luna at Standard speed in the desktop app, subject to rollout, while ChatGPT Plus ($20 a month) adds the CLI, IDE extension, web and iOS apps. Above Plus, ChatGPT Pro comes in $100, $200 and $500 monthly tiers (with Ultrafast mode on the $500 plan using 8x included usage), alongside Business (2 or more users, $20 per user a month billed yearly or $25 monthly), Enterprise and Edu plans. You can also sign in with an OpenAI API key to use the CLI, SDK and IDE extension billed at API rates, though you don't get the cloud features.

For Claude Code pricing, Claude Free doesn't include Claude Code at all, so access starts on Claude Pro at $20 a month billed monthly or $17 a month billed yearly ($200 up front). Heavy users can move up to Claude Max 5x at $100 a month or Max 20x at $200 a month, both billed monthly only and bundled with $100 or $200 a month in Claude Platform API credits. For organizations, Claude Team includes Claude Code on standard seats ($20 per seat a month billed yearly, $25 monthly) and premium seats with 5x the usage of standard seats ($100 per seat a month billed yearly, $125 monthly), while Enterprise is $20 per seat a month billed yearly with a 20-seat minimum plus usage at API rates.

Usage limits

The table below shows the Codex usage limits for local messages per 5-hour window on ChatGPT Plus and Standard Business.

Anthropic handles Claude usage limits differently, selling capacity as Pro, 5x Pro (on Max 5x) or 20x Pro (on Max 20x) with weekly limits, while Team premium seats get 5x the usage of Team standard seats. Pro and Max both include Opus, Sonnet and Haiku, while Fable runs on usage credits on Pro and is capped at up to 50% of your weekly limits on Max.

Benchmarks: Claude vs GPT-6

The table below pulls scores and per-task costs from Vals AI (updated October 29, 2026), which tests every model inside the same neutral Mini-SWE-agent harness. Because the harness is identical across runs, these numbers compare the underlying models rather than the Codex or Claude Code apps themselves.

• Claude Opus 5.5 leads both leaderboards, hitting 65.15% on Terminal-Bench 4.0 and 70.21% on SWE-bench Multimodal.
• GPT-6.1 Sol is the cheapest strong model on the board, scoring 55.05% on Terminal-Bench 4.0 at $1.72 per task.
• Claude Sonnet 5.5 nearly matches Opus 5.5 on Terminal-Bench 4.0 at 64.14%, but drops to 58.33% on SWE-bench Multimodal.

Which one to pick

• If you want to start free, pick Codex, since ChatGPT Free and Go ($8 a month) give you GPT-6 Luna in the desktop app, subject to rollout.
• If you want the top benchmark model, pick Claude Code, which gives you Claude Opus 5.5 on Pro and Max plans.
• If you want limits you can plan around, pick Codex for its published 5-hour message ranges per model on Plus and no five-hour limit on ChatGPT Pro.
• If you want an open-source CLI you can audit or fork, pick Codex, which ships under the Apache-2.0 license while Claude Code is proprietary under Anthropic's Commercial Terms of Service.
• If you work in JetBrains IDEs, pick Claude Code, which ships an official JetBrains plugin, while the Codex IDE extension covers VS Code, Cursor and Windsurf.

Use both subscriptions in other tools

Neither tool accepts a standard OpenAI-compatible Chat Completions endpoint out of the box. Custom model providers in Codex must expose the OpenAI Responses API, and Claude Code expects gateways to speak the Anthropic Messages API format (or Bedrock and Vertex formats).

Plan2API turns a Claude Pro or Max plan and a ChatGPT plan (through Codex) into one OpenAI-compatible API key at https://api.plan2api.com/v1, exposing model IDs like claude/claude-opus-5-5 and openai/gpt-6.1-sol for tools like OpenCode, while Claude Code itself can reach GPT models through Claude Code Router. Connecting both plans takes Plan2API Pro ($12 a month, which covers up to 3 connected subscriptions with no daily request cap), whereas Plan2API Free ($0) covers 1 connected subscription and 50 chat requests a day.