--- title: "Reason vs Codex: Coding Agent Index 65.1 vs 62.5 | Reason Machines" description: "Reason vs Codex on the same model: Reason 1.1 scored 65.1 against Codex's 62.5 on the Coding Agent Index v1.5 with GPT-6 Astra, at 22% lower cost per solved task. Conditions, caveats and plans." url: "https://reasonmachines.com/guides/reason-vs-codex" section: "Coding agent guides" --- # Reason vs Codex In Reason's published benchmark, Reason 1.1 and Codex ran the same model, GPT-6 Astra at xhigh effort, on the same tasks with the same graders: Reason scored 65.1 on the Coding Agent Index against Codex's 62.5, at 22% lower cost per solved task. Codex is OpenAI's agent, included in ChatGPT plans; Reason is an autonomous software engineer that works on its own cloud computers or yours and opens pull requests. Updated Nov 2026 [Start with Reason](https://reasonmachines.com/) on the web, or [download Reason for macOS](https://reasonmachines.com/download). ## Try Reason and Codex on the same task 1. **Run the task in Codex**: Give Codex a task in your repository that you already know the answer to, such as fixing a failing test. 2. **Run the same task in Reason**: With a Reason plan, open Reason on the web or in the Mac app, tell it the repository and paste the same prompt. It works on a Reason cloud computer, or on your own computer through the Mac app. 3. **Compare the diffs and the cost**: Compare whether each change is correct, how much each run cost under its plan, and how long it took. One task is an anecdote: try several. ## The benchmark and its conditions | | Reason 1.1 | Codex | | --- | --- | --- | | Coding Agent Index v1.5 | 65.1 | 62.5 | | Cost per solved task | 22% lower than Codex | Baseline | | Model and effort | GPT-6 Astra, xhigh | GPT-6 Astra, xhigh | | Source of the score | Reason's run, public Harbor job | Codex's public Harbor jobs on the same tasks | Reason's benchmark post of October 10, 2026 reports that on GPT-6 Astra, Reason beats Codex on the Coding Agent Index, 65.1 against 62.5, at 22% lower cost per solved task. The index is Coding Agent Index v1.5 (DeepSWE v1.1, Terminal-Bench 4.0 and SWE-Atlas QnA), three attempts per task. Cost per attempt is the mean over all 303 tasks, and cost per solved task divides it by the index. Caveats: this is Reason's own benchmark of its own agent, with the methodology and runner published for anyone to reproduce. On October 22, 2026, after the post, Artificial Analysis moved to Coding Agent Index v2.0, so these scores are not directly comparable with its current leaderboard. Results on other models and efforts differ. - [Reason: Reason 1.1 Coding Agent Index results (Oct 10, 2026)](https://reasonmachines.com/blog/introducing-reason-agent-1-1.md) - [Reason benchmark repository and methodology](https://github.com/reason-machines/benchmark) ## How they differ to use | | Reason | Codex | | --- | --- | --- | | Where it runs | Reason cloud computers or your own computer | Your computer (CLI, IDE, desktop), or Codex Cloud | | Models | Reason's models, plus ChatGPT, Copilot and other subscriptions or API keys you connect | OpenAI models, or open models with --oss | | Pricing | Pro $50 a month, Max $200 a month; usage in credits | Included in ChatGPT plans, from Free; Plus $20 a month | | When a limit hits | Rolls the turn over to another connected account or Reason credits | Finishes the active turn; then credits, an API key, or wait | | Apps | Web app at reasonmachines.com and the Mac app | CLI, IDE extension, desktop app, web and iOS | - [Reason docs: plans](https://docs.reasonmachines.com/docs/plans) - [OpenAI: Codex pricing and usage limits](https://learn.chatgpt.com/docs/pricing) - [Never hit a usage limit mid-task](https://reasonmachines.com/guides/usage-limit-fallback.md) - [Codex usage limit: what to do](https://reasonmachines.com/guides/codex-usage-limit.md) - [Codex vs Claude Code](https://reasonmachines.com/guides/codex-vs-claude-code.md) ## Questions ### Is Reason better than Codex? On GPT-6 Astra at xhigh, Reason 1.1 scored 65.1 against Codex's 62.5 on Coding Agent Index v1.5 in Reason's published, reproducible runs, at 22% lower cost per solved task. That is one model and one benchmark; try both on your own work. ### Can Reason use my ChatGPT subscription? Yes. Reason lets you connect a ChatGPT Plus, Pro or Business plan in Settings › Models, through Sign in with ChatGPT or Codex sign-in. Usage limits on that account come from your ChatGPT plan. ### Can I use Reason and Codex together? Yes. Connect your ChatGPT plan to Reason in Settings › Models, and carry a task from Codex on in Reason's web or Mac app with the branch and your notes. ## More guides - [Codex vs Claude Code](https://reasonmachines.com/guides/codex-vs-claude-code.md) - [Codex usage limit: what to do](https://reasonmachines.com/guides/codex-usage-limit.md) - [Never hit a usage limit mid-task](https://reasonmachines.com/guides/usage-limit-fallback.md) - [Claude Code usage limit reached: what to do](https://reasonmachines.com/guides/claude-code-usage-limit.md) - [How the Claude usage limit works](https://reasonmachines.com/guides/claude-usage-limit-explained.md) - [When the Claude weekly limit resets](https://reasonmachines.com/guides/claude-weekly-limit-reset.md) - [All coding agent guides](https://reasonmachines.com/guides.md) - [Download Reason for macOS](https://reasonmachines.com/download)