AI Compare · Coding & Development
Coding and development: which AI fits, by plan and price
Which plans document AI in your editor, completions as you type, cloud agents that open pull requests and automatic code review, what each costs, and how usage is counted.
Prices for United States, before sales tax Data release 2026.10.16-3, Oct 16, 2026 Rules 2.1.0
On AI Compare, coding covers five tasks: getting help with code questions, diagnosing and repairing a small program, coding inside your editor, handing repository tasks to an agent, and reviewing pull requests automatically. We compare the coding side of four general assistants, ChatGPT (Codex), Claude (Claude Code), Gemini (Jules and Google Antigravity) and Perplexity, with two coding products, Cursor and GitHub Copilot.
The facts here come from each company's own pricing and help pages for individual US plans, checked on October 12 and 13, 2026. They tell you which plans document each of these tools and how usage is metered. They cannot tell you how often the code is correct or how useful a review is. A listed feature is not proof that it works well, and we have not run our tests on these plans yet.
The tasks, and what each one needs
- Get help with code questions Explain errors, write snippets and answer programming questions.
- Diagnose and repair a small program Find the bugs in a short program and fix it without breaking anything else.
- Code inside my editor Get completions and edits inside the code editor you already use.
- Hand repository tasks to an agent Give an agent an issue and get a pull request back for review.
- Review pull requests automatically Get review comments on pull requests before a person reviews them.
- Build an app without writing code Describe an app or website and get a working version you can change by chatting.
- Put an app or site online Publish what you built to a live web address, on your own domain if you want.
Which plans document each task
For each product, the lowest-priced plan that documents every feature the task needs. Documented means offered, not tested.
| Product | Get help with code questions | Diagnose and repair a small program | Code inside my editor | Hand repository tasks to an agent | Review pull requests automatically | Build an app without writing code | Put an app or site online |
|---|---|---|---|---|---|---|---|
| ChatGPT | ChatGPT FreeFree | ChatGPT FreeFree | ChatGPT Plus$20 a month | ChatGPT Plus$20 a month | ChatGPT Plus$20 a month | ChatGPT Plus$20 a month | ChatGPT Plus$20 a month |
| Claude | Claude FreeFree | Claude FreeFree | Claude Pro$20 a month | Claude Pro$20 a month | Claude Pro$20 a month | Claude FreeFree | Claude Pro$20 a month |
| Gemini | Gemini (free)Free | Gemini (free)Free | Gemini (free)Free | Gemini (free)Free | Not offered | Gemini (free)Free | Gemini (free)Free |
| Perplexity | Perplexity FreeFree | Perplexity FreeFree | Perplexity Pro$20 a month | Not verified | Not verified | Perplexity Pro$20 a month | Perplexity Pro$20 a month |
| X Subscription (Cursor, Grok, X) | X Subscription FreeFree | X Subscription FreeFree | X Subscription FreeFree | X Subscription Premium$30 a month | X Subscription Premium$30 a month | X Subscription FreeFree | X Subscription FreeFree |
| GitHub Copilot | Copilot FreeFree | Copilot FreeFree | Copilot FreeFree | Copilot Pro$10 a month | Copilot Pro$10 a month | Not verified | Not verified |
| Suno | Suno FreeFree | Suno FreeFree | Not offered | Not offered | Not offered | Not offered | Not offered |
| Udio | Udio FreeFree | Udio FreeFree | Not offered | Not offered | Not offered | Not offered | Not offered |
| ElevenLabs | ElevenLabs FreeFree | ElevenLabs FreeFree | Not offered | Not offered | Not offered | Not offered | Not offered |
| Otter | Otter BasicFree | Otter BasicFree | Not offered | Not offered | Not offered | Not offered | Not offered |
| Wispr Flow | Flow FreeFree | Flow FreeFree | Not offered | Not offered | Not offered | Not offered | Not offered |
| Superwhisper | Superwhisper FreeFree | Superwhisper FreeFree | Not offered | Not offered | Not offered | Not offered | Not offered |
| Runway | Runway FreeFree | Runway FreeFree | Not offered | Not offered | Not offered | Not offered | Not offered |
| Luma | Luma Plus$30 a month | Luma Plus$30 a month | Not offered | Not offered | Not offered | Not offered | Not offered |
| Higgsfield | Higgsfield FreeFree | Higgsfield FreeFree | Not offered | Not offered | Not offered | Not offered | Not offered |
| HeyGen | HeyGen FreeFree | HeyGen FreeFree | Not offered | Not offered | Not offered | Not offered | Not offered |
| Synthesia | Synthesia BasicFree | Synthesia BasicFree | Not offered | Not offered | Not offered | Not offered | Not offered |
| Descript | Descript FreeFree | Descript FreeFree | Not offered | Not offered | Not offered | Not offered | Not offered |
| Midjourney | Midjourney Basic$10 a month | Midjourney Basic$10 a month | Not offered | Not offered | Not offered | Not offered | Not offered |
| Ideogram | Ideogram FreeFree | Ideogram FreeFree | Not offered | Not offered | Not offered | Not offered | Not offered |
| Adobe Firefly | Firefly FreeFree | Firefly FreeFree | Not offered | Not offered | Not offered | Not offered | Not offered |
| Canva | Canva FreeFree | Canva FreeFree | Not offered | Not offered | Not offered | Canva Pro$18 a month | Canva FreeFree |
| Lovable | Lovable FreeFree | Lovable FreeFree | Lovable Pro$25 a month | Not offered | Not offered | Lovable FreeFree | Lovable FreeFree |
| Replit | Replit StarterFree | Replit StarterFree | Replit StarterFree | Not offered | Not offered | Replit StarterFree | Replit StarterFree |
| v0 | v0 FreeFree | v0 FreeFree | v0 FreeFree | Not offered | Not offered | v0 FreeFree | v0 FreeFree |
| Microsoft Copilot | Microsoft Copilot (free)Free | Microsoft Copilot (free)Free | Not verified | Not verified | Not verified | Microsoft 365 Personal$9.99 a month | Not verified |
Test them on your own work
Documented features say what a plan offers, not how well it does the work. Run the same kit in each AI you are considering and score the answers, blind if you like.
-
Diagnose and repair a small program
A twenty-line Python function with three real bugs and six unit tests. Run the tests on your own computer: this site never runs code.
-
Implement a small specification
A written specification for a Python function, seven visible tests and five held-out tests you keep to yourself. Held-out tests show whether the code follows the specification or only the tests it was shown.
-
Refactor code without changing what it does
A working but deeply nested shipping-cost function, seven tests to keep passing and nine held-out tests that pin rounding, error order and edge cases. A refactor that changes behaviour fails, however clean it looks.
How to choose
For code questions and small repairs, any coding agent or chat can start. ChatGPT, Gemini, Cursor and GitHub Copilot document a coding agent on their free tiers, with limited use. Claude includes Claude Code from Pro ($20 a month, or $200 a year paid up front); Claude Free still has chat and code execution. Perplexity lists no dedicated coding agent. Our code-repair kit gives each assistant the same short Python function with bugs and a test file you must not change, so you can run the tests yourself and see whether the fix holds.
To code inside your editor, check which editors and whether completions are documented.
- GitHub Copilot: extensions for VS Code, Visual Studio, JetBrains IDEs, Xcode, Eclipse, Neovim and Zed on every plan. Completions as you type are capped at 2,000 a month on Free and unlimited, outside AI credits, on Pro ($10 a month), Pro+ ($39) and Max ($100).
- Cursor: its own editor on every plan, with its agent also in JetBrains IDEs and Xcode. Tab completions are documented from Premium ($30 a month) up; for Free and Plus ($8) the Tab limit is not stated, so that is not verified.
- Gemini: Google Antigravity is free with a personal Google account, as an IDE, a CLI and extensions for VS Code, Visual Studio, JetBrains, Zed and Xcode. Unlimited Tab completions are documented in the Antigravity IDE on every plan; for the extensions this is not stated.
- ChatGPT: the Codex extension for VS Code, Cursor and Windsurf, with Xcode and JetBrains through their own integrations, on Plus ($20) and Pro ($100, $200 or $500). OpenAI's pages conflict for Free and Go. Completions as you type are not documented on any plan.
- Claude: Claude Code for VS Code, Cursor and JetBrains IDEs on Pro and Max ($100 or $200), not on Free. Completions as you type are not documented, and Anthropic's desktop app docs say it has none.
- Perplexity: no in-editor assistant. Pro and Max can connect an editor to Perplexity Computer, which takes delegated tasks billed in credits.
To hand repository work to an agent, look for a cloud agent that opens pull requests.
- Gemini: Jules (beta, 18+) on every plan, free included: 15 tasks a day and 3 at once on Free, 100 and 15 on Google AI Pro ($19.99 a month, or $199.99 a year), 300 and 60 on Ultra. It works on connected GitHub repos.
- GitHub Copilot: cloud agent on Pro, Pro+ and Max, for GitHub-hosted repos only, one pull request per task, with a 59-minute session limit. Pro+ and Max can also hand tasks to third-party agents (Claude, Codex).
- Claude: Claude Code on the web on Pro and Max, in Anthropic-managed cloud machines that clone GitHub repos and create pull requests, within your plan's usage limits.
- ChatGPT: Codex Cloud on Plus and Pro, subject to rollout; OpenAI says it is not included with Free or Go.
- Cursor: Cloud Agents on Premium, Super ($100) and Ultra ($200), for GitHub, GitLab, Bitbucket and Azure DevOps repos; not on Free or Plus.
- Perplexity: Computer runs code in a cloud sandbox and connects to GitHub, but no page we checked says it opens pull requests, so that is not verified.
For automatic pull request review, fewer plans qualify. GitHub Copilot reviews pull requests on Pro and up, at roughly $0.05 to $5 of AI credits per review depending on the mode; Free only reviews a selection in VS Code. Cursor's Bugbot reviews GitHub and GitLab pull requests from Premium up, billed by usage. ChatGPT Plus and Pro include Codex code review on GitHub; for Free and Go it is not stated. Claude Pro and Max can review through Claude Code GitHub Actions with your subscription; Anthropic's managed Code Review is for Team and Enterprise plans only. Google documents no pull request reviewer for consumer Gemini plans since it shut down Gemini Code Assist on GitHub on July 17, 2026. Perplexity does not document one.
Compare how usage is counted, not only the price. Copilot meters chat, agents and review in AI credits worth one cent each: 1,500 a month on Pro, 7,000 on Pro+ and 20,000 on Max, with completions outside credits. Claude's chat and Claude Code share one pool with a five-hour session limit and a weekly limit. On ChatGPT, Codex shares Plus usage limits, and OpenAI says the Pro plans have no five-hour limit for now. Cursor's plans, sold since October 12, 2026 as the X Subscription with Grok and X Premium, share one weekly pool with Grok; models from Anthropic, OpenAI and Google cost extra at API list price. Claude Pro and Google AI Pro can be paid yearly, up front; Copilot, Cursor and ChatGPT are monthly for new subscribers. All prices are US list prices before tax.
Check what happens to your code. GitHub says it may train on Copilot interaction data from individual plans by default since April 24, 2026; you can opt out in settings. Cursor's Privacy Mode keeps your code out of training, and with it off Cursor may train on it. ChatGPT trains by default until you turn off "Improve the model for everyone", Gemini until you turn off Keep Activity, and Anthropic does not state Claude's default. We have no kits yet for editor work, repository agents or pull request review, so try them on a test repository first.
Questions
Can I code with AI for free?
Yes, with limits. Copilot Free gives 2,000 completions a month and limited chat and agent use. Google Antigravity is free with a Google account and documents Tab completions in its IDE. Gemini's Jules and the free tiers of ChatGPT and Cursor include limited agent use. Claude Code needs Pro.
Which plans include an agent that opens pull requests?
Documented: Jules on every Gemini plan, Copilot cloud agent on Pro and up, Claude Code on the web on Pro and Max, Codex Cloud on ChatGPT Plus and Pro, and Cursor Cloud Agents from Premium. Perplexity does not document opening pull requests.
Do I have to switch editors?
Not necessarily. Copilot, Claude Code, Codex and Google Antigravity all document extensions for VS Code and support for JetBrains IDEs. Cursor is its own editor, though its agent also runs in JetBrains IDEs.
Will my code be used to train models?
It depends on your settings. GitHub may train on Copilot interaction data by default, with an opt-out. Cursor's Privacy Mode keeps code out of training. ChatGPT and Gemini train by default until you switch it off, and Anthropic does not state Claude's default.
Do these features mean the code will be correct?
No. They show what each plan offers, not how often the code works. Our code-repair kit gives every assistant the same buggy function and test file, so you can run the tests yourself. We have not published results yet.
Written by the What the tech desk with AI assistance from the facts in our data release 2026.10.16-3. Read how AI Compare works.