GitHub Copilot Review 2026: Plans, AI Credits and Who It Suits
Disclosure: GitHub does not pay us, and the GitHub links on this page are not affiliate links. The three Amazon links near the end are. Our affiliate disclosure explains how the site is funded.
The short version: GitHub Copilot is the easiest AI coding assistant to recommend if you want it inside the editor you already use. It runs in VS Code, Visual Studio, JetBrains, Vim and Neovim, Xcode and Eclipse, and completions are unlimited on every paid plan. Pro costs $10 a month. Two things changed in 2026 that older reviews miss. Usage beyond completions is now measured in AI Credits, and since April 24, 2026 GitHub uses Free, Pro and Pro+ interaction data to train its models unless you opt out.
How we reviewed it
We did not run Copilot on our own projects, and we do not publish scores or accuracy percentages of our own. Everything below comes from GitHub’s own plans pages, documentation and changelog, read on October 27, 2026.
What GitHub Copilot does in 2026
- Completions and next edit suggestions: the original grey ghost text, plus predictions of your next edit elsewhere in the file. Both are unlimited on paid plans (Copilot plans).
- Copilot Chat and agent mode: chat in the IDE and on GitHub.com, and an agent mode that edits across files in your editor. Both are on every plan, Free included, with Free usage limited.
- Copilot cloud agent: formerly called the Copilot coding agent. You assign it an issue or a pull request, it works in the background and opens one pull request per task, with a 59-minute session limit (cloud agent docs). It is included from Pro up.
- Copilot code review: automated review comments on pull requests, from Pro up. Free gets only “Review selection” in VS Code.
- Copilot CLI: a terminal agent. The old
gh copilot suggestextension stopped working on October 25, 2025. Since January 21, 2026 thegh copilotcommand installs and launches the new CLI (GitHub changelog). - Copilot Spaces: shared bundles of repositories, issues, notes and files that give Chat context. Available on every licence, including Free (Spaces docs).
If you read about Copilot Workspace elsewhere, that technical preview was retired on May 30, 2025. Its job is now done by the cloud agent.
GitHub Copilot pricing (October 2026)
| Plan | Price | AI Credits a month | Completions | Cloud agent and code review |
|---|---|---|---|---|
| Free | $0 | Limited chat and agent use | 2,000 a month | No cloud agent; review selection only |
| Pro | $10/mo | 1,500 | Unlimited | Included |
| Pro+ | $39/mo | 7,000 | Unlimited | Included |
| Max | $100/mo | 20,000 | Unlimited | Included |
| Business | $19/seat/mo | 1,900 | Unlimited | Included |
| Enterprise | $39/seat/mo | 3,900 | Unlimited | Included |
Sources: GitHub’s Copilot plans documentation and plans page, read October 27, 2026. One AI Credit is worth $0.01.
The model is simple once you see it. Completions are unlimited on any paid plan, and credits pay for everything conversational: chat, agent mode, the cloud agent and code review. Heavier models use more credits per request. If most of your use is autocomplete, Pro at $10 goes a long way. If you hand the cloud agent several tasks a day, look at Pro+ or Max.
Models
Copilot’s model menu is one of the broadest available. GitHub’s supported-models reference lists OpenAI’s GPT-5 and GPT-6 families, Anthropic’s Claude models up to Opus 5.5 and Fable 5.1, Google’s Gemini 3.8 Flash and Gemini 4 Argon, xAI’s Grok, Moonshot’s Kimi K3 and Microsoft’s MAI-Code-1.1-Flash. Chat defaults to Auto selection, which picks a model for you. GitHub does not name a fixed default for completions. The Free plan’s documentation says “Auto model selection only”, although GitHub’s marketing page lists a wider set, so expect Free to be steered to Auto.
Editor support
This is still Copilot’s clearest advantage. GitHub’s installation docs list VS Code, Visual Studio 2022 (built in from 17.10), JetBrains IDEs, Vim 9 and Neovim 0.6 or later through copilot.vim, Xcode, Eclipse 2024-03 or later and Azure Data Studio. A team split between VS Code, IntelliJ and Vim can standardise on Copilot without asking anyone to change editors. That is not true of Cursor or Devin Desktop.
Privacy and training
This changed this year, and it matters more than any feature. GitHub’s March 25, 2026 changelog says that from April 24, 2026, interaction data from Copilot Free, Pro and Pro+ is used “to train and improve our AI models unless you opt out” (GitHub changelog). The switch is under Privacy at github.com/settings/copilot, and earlier opt-outs carry over. Business and Enterprise are not affected. They are covered by GitHub’s Data Protection Agreement, under which your data is not used for AI training.
If you work on client or employer code on a personal Pro account, turn the setting off before you start, or ask for a Business seat.
What we like
- The widest editor support in the category, Vim and Xcode included
- Unlimited completions on every paid plan from $10
- A cloud agent and code review built into GitHub itself, from Pro
- A usable Free plan with 2,000 completions a month
- A very broad model menu across OpenAI, Anthropic, Google and others
What could be better
- Individual plans train on your interaction data unless you opt out
- Credits make agent-heavy use harder to predict than a flat fee
- Free is limited to Auto model selection, per GitHub’s docs
- GitHub’s own pages disagree on some Free-plan details
Copilot vs the alternatives
vs Cursor: Cursor is a whole editor built around its agent, with Grok models on its $8 Plus plan and third-party models from $30 plus API-priced usage. Copilot is cheaper for autocomplete, works in more editors, and does not train on Business or Enterprise data. Cursor suits developers who want to switch editors for a deeper agent. See our Cursor review.
vs Claude Code and Codex: these coding agents come with $20 chat plans. They are stronger for handing over whole tasks, but Claude Code offers no inline autocomplete. Many developers run Copilot for completions alongside one of them. See our Claude vs ChatGPT for coding comparison.
For all six tools side by side, see our best AI coding assistants roundup.
Who should use GitHub Copilot?
A strong fit: teams on mixed editors, JetBrains or Vim users, anyone whose code already lives on GitHub and who wants the agent to open pull requests there, and organisations that need a no-training data agreement (Business or Enterprise).
Look elsewhere if: you want an editor designed around the agent (Cursor or Devin Desktop), or you already pay for Claude Pro or ChatGPT Plus and mainly want an agent rather than autocomplete.
Frequently asked questions
Is GitHub Copilot free?
Yes, there is a Free plan with 2,000 completions a month and limited chat and agent-mode use. It does not include the cloud agent.
What are GitHub AI Credits?
The unit Copilot now uses for chat, agents and code review. One credit is worth $0.01. Pro includes 1,500 a month, Pro+ 7,000 and Max 20,000. Completions do not use credits on paid plans.
Does GitHub Copilot train on my code?
On Free, Pro and Pro+, GitHub uses interaction data for training unless you opt out in your Copilot settings, effective April 24, 2026. Business and Enterprise data is not used for training.
What happened to Copilot Workspace?
The preview ended on May 30, 2025. The Copilot cloud agent, which turns an issue into a pull request, now covers that job.
Does Copilot work in Vim and Neovim?
Yes, through the copilot.vim plugin, on Vim 9.0.0185 or later and Neovim 0.6 or later.
Our verdict
GitHub Copilot is the safest default for AI help inside an existing editor. It has the broadest IDE support, unlimited completions from $10, and an agent and code reviewer that work where your code already lives. Its weak points are new. Credits make heavy agent use less predictable, and individual plans now train on your interactions unless you switch that off. Start on Free, move to Pro when you hit the completion cap, and check the training setting on day one.
About the author
Lennit Herrera is the founder and editor of AI Tool Stack Reviews, and he builds and audits websites for a living. We do not run hands-on tests and we do not publish scores of our own. Prices, limits and features come from the vendor’s own pages and are dated. Corrections are welcome through our contact page.
Gear we’d pair with Copilot
- A solid mechanical keyboard. The better your typing setup, the more autocomplete pays off.
- An ergonomic chair, because you’ll be at the keyboard longer than you think.
- Clean Code. AI drafts it; you still have to keep it maintainable.
As an Amazon Associate we earn from qualifying purchases.
