Skip to main content

Codex vs GitHub Copilot: ChatGPT-Bundled Agent or GitHub-Native Platform?

Choose Copilot when GitHub should govern agent work and its budget, and Codex for OpenAI's current models in Codex Cloud. In public studies, Codex's agent pull requests are merged at a higher rate.

Reading time
3 min read
Sources
19 cited
In this piece

GILD verdict

Choose Codex when work should run on OpenAI's current models in hosted Codex Cloud environments, started from ChatGPT, the CLI, Slack or Linear.

Choose Copilot if GitHub is your system of record and you want agent work and its budgets governed there. Copilot can still run the Codex agent, on the models GitHub lists. Public merge data favors Codex, whose agent pull requests merged at an estimated 73.5% against Copilot's 59.6% in a July 2026 study of 489 Python repositories. Pilot both on the same repositories and compare accepted, reviewed changes and total spend at the plan you would buy. Billed annually a ChatGPT Business seat costs US$20 a month, US$1 more than Copilot Business at US$19, and billed monthly the gap is US$6.

Pick by which subscription should pay for agent work and which console should govern it. Copilot is GitHub's assistant across IDEs, GitHub.com, a CLI and its own app, and its paid plans can run the OpenAI Codex agent on issues and pull requests. Codex bought through ChatGPT adds the desktop app and Codex Cloud, on OpenAI's current models.

Choosing between Codex and GitHub Copilot

Choose Codex if

  • Engineers work from the terminal, the ChatGPT desktop app or an IDE extension, and agent work need not start in GitHub.
  • You want OpenAI's current models, such as GPT-6.1 Sol, which ChatGPT plans offer in Codex.
  • You want hosted Codex Cloud environments with their own internet-access settings, started from ChatGPT, the CLI, Slack, or Linear.

Choose GitHub Copilot if

  • Your code, issues, and pull requests already live on GitHub and you want agent work assigned there.
  • You want models from several providers, and third-party Claude and Codex agents, under one GitHub bill.
  • You want pooled AI credits with budgets set in GitHub at the user, cost-center, organization, and enterprise level.

Side by side

DimensionCodexGitHub Copilot
Where it runsChatGPT desktop app, CLI, IDE extension for VS Code, Cursor, and Windsurf, and Codex Cloud45VS Code, Visual Studio, JetBrains IDEs, Eclipse, Xcode, GitHub.com, GitHub Mobile, Copilot CLI, and the Copilot app11
Individual priceList prices in US dollars. Included usage differs by plan and is not directly comparable.Included in ChatGPT Plus: US$20 per month1Copilot Pro: US$10 per month; Pro+: US$39 per month9
Team priceIncluded in ChatGPT Business: US$25 per user per month billed monthly, or US$20 billed annually (2+ users)1Copilot Business: US$19 per granted seat per month; Enterprise: US$39, with GitHub Enterprise Cloud9
How usage is meteredPlan limits that vary by model and task, then ChatGPT credits; an API key bills at API rates1AI credits pooled per organization (1,900 per Business seat, 3,900 per Enterprise seat each month), then US$0.01 per credit; completions stay unlimited10
Codex modelsGPT-6.1 Sol, GPT-6 Sol, and GPT-6 Luna on Plus, Pro, Business, and Enterprise2The Codex agent on GitHub offers GPT-5.3-Codex, GPT-5.4, and GPT-5.4 nano14
Cloud agentCodex Cloud tasks in OpenAI-managed environments; off by default on Enterprise until an admin enables it56Copilot cloud agent and third-party agents in an ephemeral environment, using GitHub Actions minutes and AI credits1213
Pull request review@codex review on connected GitHub repositories, flagging P0 and P1 issues and metered as Code Review usage18Copilot code review, about US$0.05 to US$1 per review at Lite effort or US$0.25 to US$5 at Balanced, plus Actions minutes; GitHub says 71% of reviews surface actionable feedback1517
Agent pull requests merged73.5%, estimated across 489 open-source Python repositories in a July 2026 study; 64% across the 2025 AIDev dataset181959.6% and 35% in the same two studies1819
Median wait for an accepted agent pull request18 minutes in the 2025 AIDev dataset, against 3.9 hours for human pull requests1917.2 hours in the same dataset19
Project instructionsAGENTS.md, layered from the repository root down7.github/copilot-instructions.md, path-specific instructions, and AGENTS.md16

What practitioners report, side by side

Each product's figures come from its own practitioners, for the group you pick.

CodexOpenAI+57 Strong positiveHow this figure is made
GitHub CopilotGitHub+11 MixedHow this figure is made
Whose sources to count
Compare Codex with
  • GitHub Copilot

Overall signal

Leaders and teams rate Codex +57, 46 points above GitHub Copilot at +11. Most of the gap is in praise, with 24 of Codex's 39 sources positive against 9 of GitHub Copilot's 32. Codex has 3 negative and GitHub Copilot 8.

The score counts the sources from leaders and teams that take a side and is how far positive votes lead negative ones, per 100 votes. A confirmed leader casts three votes, another confirmed practitioner two and any other source one, so Codex's +57 rests on votes split 41 positive, 17 mixed and 5 negative and GitHub Copilot's +11 on 16, 27 and 10. A lead of a fifth of the votes reads positive and a lead of half strong positive, with the same lines below zero. The dots show each source once, positive, mixed, negative. Use a task pilot to measure accepted work and review time.

Footprint on the Agentic Software Development stack

They differ on five of 12 layers. Codex owns two that GitHub Copilot does not, and GitHub Copilot owns one that Codex does not. Coverage is each product's placement on this stack, as its product page shows it. Core is a job the product owns; Supports and Touches are jobs it helps with. Coverage says nothing about how well a product does them.

Layers where Codex and GitHub Copilot differ
LayerCodexGitHub Copilot
Build
Codebase contextCoreSupports
Tool executionCoreSupports
Validate
Code reviewSupportsCore
Application securityTouchesSupports
Ship
Source controlTouchesSupports
The 7 layers where they match
Layers where Codex and GitHub Copilot match
LayerCodexGitHub Copilot
Plan
Work planningSupportsSupports
Build
CodingCoreCore
Agent runtimeCoreCore
Validate
DebuggingSupportsSupports
TestingSupportsSupports
Ship
CI/CDTouchesTouches
Operate
ObservabilityTouchesTouches

What practitioners praise

Among leaders and teams, Code Quality and Developer Experience reach 5 sources from 3 authors only for Codex, at 7 of 8 and 7 of 13. GitHub Copilot has 0 of 4 on Code Quality so far and needs a pilot on the same repository tasks before you choose on Developer Experience. Enterprise Administration reaches 5 sources from 3 authors only for GitHub Copilot, at 3 of 7. Codex needs a pilot on the same repository tasks before you choose on it.

What practitioners praise, Leaders and teams
Praised forCodexGitHub Copilot
Code Quality7 of 80 of 4 so far
Developer Experience7 of 13Not a theme in its sources
Enterprise AdministrationNot a theme in its sources3 of 7

Each count is the sources that praise a theme, or raise it as a concern, out of those that discuss it. A row pairs the two products when both carry a theme under the same or an equivalent name. A count ending in "so far" has under 5 sources from 3 authors among leaders and teams. "Not a theme in its sources" means GILD found no theme of that name in that product's sources.

What practitioners raise as concerns

GitHub Copilot's sources raise Requires Strong Review more often than Codex's. It is raised in 1 of the 9 Codex sources that discuss it, against 5 of 7 for GitHub Copilot. Speed reaches 5 sources from 3 authors only for Codex, at 3 of 5. GitHub Copilot needs a pilot on the same repository tasks before you choose on it. GitHub Copilot alone has three concern themes that reach 5 sources from 3 authors. Measuring Impact and Quality Against Alternatives lead, at 5 of 13 and 3 of 7.

What practitioners raise as concerns, Leaders and teams
Raised as a concernCodexGitHub Copilot
Usage Limits / Cost7 of 133 of 7
Requires Strong Review1 of 95 of 7
Speed3 of 5Not a theme in its sources
Measuring ImpactNot a theme in its sources5 of 13
Quality Against AlternativesNot a theme in its sources3 of 7
Data, IP & SecurityNot a theme in its sources2 of 6
Codex SignalsGitHub Copilot Signals

19 sources

  1. Codex docs: PricingOfficial documentationlearn.chatgpt.com
  2. Codex docs: ModelsOfficial documentationlearn.chatgpt.com
  3. Codex docs: AuthenticationOfficial documentationlearn.chatgpt.com
  4. Codex docs: IDE extensionOfficial documentationlearn.chatgpt.com
  5. Codex docs: CloudOfficial documentationlearn.chatgpt.com
  6. Codex docs: Admin rollout guideOfficial documentationlearn.chatgpt.com
  7. Codex docs: AGENTS.mdOfficial documentationlearn.chatgpt.com
  8. Codex docs: GitHub code reviewOfficial documentationlearn.chatgpt.com
  9. GitHub Docs: Plans for GitHub CopilotOfficial documentationdocs.github.com
  10. GitHub Docs: Usage-based billing for organizations and enterprisesOfficial documentationdocs.github.com
  11. GitHub Docs: Where to use GitHub CopilotOfficial documentationdocs.github.com
  12. GitHub Docs: GitHub Copilot on GitHub.comOfficial documentationdocs.github.com
  13. GitHub Docs: About third-party coding agentsOfficial documentationdocs.github.com
  14. GitHub Docs: OpenAI CodexOfficial documentationdocs.github.com
  15. GitHub Docs: About GitHub Copilot code reviewOfficial documentationdocs.github.com
  16. GitHub Docs: Adding repository custom instructionsOfficial documentationdocs.github.com
  17. GitHub Blog: 60 million Copilot code reviews and counting (Mar 5, 2026)Public sourcegithub.blog
  18. arXiv: How Do AI Coding Agents Contribute to Software Development? An Empirical Study of Agentic Pull Requests (Jul 23, 2026)Public sourcearxiv.org
  19. arXiv: The Rise of AI Teammates in Software Engineering 3.0, the AIDev dataset (Jul 2025)Public sourcearxiv.org
Read the research