Table of Contents
GitHub Copilot pricing 2026: Pro vs Pro+ vs Max, which plan covers your usage
On June 1, 2026, GitHub flipped every Copilot plan from premium request units to token-based GitHub AI Credits, so your monthly bill now tracks the actual tokens your chats and agent sessions burn (GitHub Changelog, June 1). That kills the old "one flat fee, use it however" math and replaces it with a real question: does your daily work fit inside Pro's allowance, or are you about to blow through it by Tuesday? If you missed the switch itself, see our June 1 note. This guide is the part that comes after, picking the plan whose credit pool covers the way you actually code. The plans page lives at github.com/features/copilot.
The one fact that changes how you read every plan: code completions and Next Edit suggestions are unlimited on every paid plan and do not touch your credits (GitHub Docs). Tab-complete all day, none of it draws down the balance. Credits get spent by chat, Copilot CLI, the cloud agent, Spaces, Spark, and third-party coding agents. So the plan you need is a function of how much agentic work you do, not how much you type.
How AI Credits actually work
One AI credit equals $0.01 USD (GitHub Docs). Every interaction consumes input, output, and cached tokens, each priced per the model you picked, and that token cost converts straight into credits. A quick question to a lightweight model costs a fraction of a credit. A long cloud agent run on a frontier model costs real money, because it is doing real work.
Each paid plan splits its monthly allowance into two parts (GitHub Docs):
- Base credits. Included with your subscription, matched to the price, and they never change.
- Flex allotment. An extra monthly amount on top, deliberately variable, designed to move as model pricing and efficiency change.
Base credits get spent first; the flex allotment kicks in automatically across your IDE, GitHub.com, and the CLI with no setup. The catch: the flex number is not a promise you can budget against month to month. GitHub built it to flex, so treat your base credits as the floor you can count on and the flex as headroom that may shift.
The four individual plans at a glance
Here are the current individual plans, with figures pulled from GitHub's billing docs on 2026-06-11. The "estimated agentic sessions" column is ours, derived from the published per-token rate table (more on the math below); read it as a planning rough-cut, not a guarantee.
| Plan | Monthly price | Included AI credits (GitHub Docs) | Estimated agentic sessions/month | Best for |
|---|---|---|---|---|
| Free | $0 | Credit allowance + 2,000 completions/mo | A handful of light agent runs | Trying Copilot, hobby repos, unlimited tab-complete |
| Pro | $10 | 1,500 (1,000 base + 500 flex) | Roughly 10 to 25 medium sessions | Solo devs living mostly in chat and completions |
| Pro+ | $39 | 7,000 (3,900 base + 3,100 flex) | Roughly 50 to 120 medium sessions | Engineers running cloud agents on feature branches |
| Max | $100 | 20,000 (10,000 base + 10,000 flex) | Roughly 150 to 350 medium sessions | CLI fleet mode and security review on every PR |
Prices and credit allowances per GitHub Docs; Free completions count per the same page. Two things to flag before you read this as gospel. Max is an upgrade-only tier right now, open to existing Student, Pro, and Pro+ subscribers, and new sign-ups for all of these plans are still paused as of June 1 (GitHub Changelog, June 1). And the session counts swing wide on purpose, because a session's cost depends entirely on the model and how much context it chews.
Where the session estimate comes from
Nobody can hand you a fixed per-session number, because the cost moves with the model. So here is the rubric, anchored to GitHub's published rates so you can run it for your own model.
Take Claude Sonnet 4.6, a common pick for agent work: $3.00 per million input tokens and $15.00 per million output (GitHub models and pricing). Say a medium agent session reads roughly 200K tokens of context and writes 50K of edits and explanation. That is 0.2 x $3.00 plus 0.05 x $15.00, so about $1.35, or 135 credits.
Example: at ~135 credits per medium Sonnet 4.6 session, Pro's 1,500 credits cover roughly 11 sessions before you touch the flex headroom; Pro+'s 7,000 cover around 50. Reach for a frontier model like Claude Fable 5 at $10 input / $50 output (same rate table) and a single heavy session can run several hundred credits on its own.
That spread is why the table's session column is a range, not a point. Light chat in a cheap model stretches a plan a long way. A frontier-model agent grinding across a monorepo eats it fast.
Pick your plan by how you actually work
Three shapes cover most of who lands on this page. Find yours.
The solo dev who lives in chat and completions
You write code most of the day with tab-complete, ask Copilot Chat to explain a stack trace or sketch a function, and occasionally kick off a small agent task. You are not running autonomous multi-file refactors on the regular.
Completions are free and unlimited, so the bulk of your usage never costs a credit (GitHub Docs). Your credit draw is mostly chat, and a quick question on a lightweight model is a fraction of a credit. Pro's 1,500 credits absorb a steady chat habit plus a dozen-ish small agent runs without you ever thinking about the balance. One lever to know: on a paid plan you get a 10% discount on model costs when you use auto model selection in Chat, the CLI, or the cloud agent (GitHub Docs). Leave it on auto and the cheap models handle the cheap questions.
Verdict: Pro at $10/month (GitHub Docs). It is the floor for anyone past the free tier, and a chat-and-completions workflow rarely escapes its allowance. Don't pay for Pro+ headroom you won't spend. Current pricing is on the plans page.
The mid-level engineer running cloud agents on feature branches
Your day includes real agentic work. You hand the cloud agent a branch, let it implement a feature and open the PR, and you do this several times a week. With the June 10 update, Copilot Chat can pull in your agent session logs and search past sessions, so you ask follow-ups on what the agent changed without leaving the conversation (GitHub Changelog, June 10). Useful, and also a reminder: every one of those agent calls spends credits.
Run the rubric. At roughly 135 credits for a medium Sonnet session, Pro's 1,500 cover only about ten before you hit the flex edge, and a busy feature week burns that in days. Pro+'s 7,000 credits (GitHub Docs) give you real room: somewhere near 50 sessions at that model, more if you keep auto selection on. This is the tier the agentic workflow was priced for.
Verdict: Pro+ at $39/month (GitHub Docs). If cloud agent sessions are a routine part of shipping features, Pro will have you watching the meter; Pro+ lets you forget it most months. The upgrade path runs through your billing settings on the plans page.
The power user with CLI fleet mode and security review on every PR
You run Copilot CLI in fleet mode, executing tasks in parallel, and you wired up the security-review command so it scans every pull request. This is the heaviest individual profile there is. Parallel CLI tasks mean several agent sessions running at once, each drawing credits, and PR-level review adds a model pass on every push. On top of credits, Copilot code review now also consumes GitHub Actions minutes at standard Actions rates (GitHub Changelog, June 1), so your spend has two meters running, not one.
Pro+ at 7,000 credits covers a moderate version of this, but fleet mode plus per-PR review is exactly the workload that empties it before month-end. Max carries 20,000 credits and higher spending limits built for intensive workflows (GitHub Changelog, June 1). At the Sonnet rubric rate that is a couple hundred medium sessions, fewer on heavier models, with the headroom to run the fleet without throttling yourself.
Verdict: Max at $100/month, if you can get it. It is the only tier sized for fleet mode plus continuous review, but it is upgrade-only today, available to existing Student, Pro, and Pro+ subscribers while new sign-ups stay paused (GitHub Changelog, June 1). If you are already on Pro or Pro+ and slamming the ceiling, upgrade from your billing settings. If you are not in yet, sit on Pro+ until sign-ups reopen.
For team leads: stop a runaway session from eating the month
Different problem if you manage seats. The failure mode that bit people on June 1 was an agent session running long and quietly burning a month's allocation in an afternoon. Two controls landed alongside the billing switch to prevent that.
First, user-level budgets, now generally available for organizations and enterprises (GitHub Changelog, June 1). Admins set a universal budget across users or override it for specific groups. The part that matters: for AI credits, a user-level budget controls total usage, not just the spend past the included pool. So you can cap a seat's entire monthly consumption, not merely the overage. As users near their cap, admins get email and can bump the budget from billing settings on the fly. That is your hard stop against one runaway session draining the org. Set those ceilings before a heavy week, not after the invoice; the flow lives in GitHub's budget management docs.
Second, the default Actions runner for code review. Since June 1, code review consumes Actions minutes on top of credits, and by default it uses a standard GitHub-hosted runner. Org admins can now set a default runner across every repository at once, instead of configuring each repo by hand (GitHub Changelog, June 1). If you run a lot of automated review, point that default at the runner class whose per-minute rate you want to pay, org-wide, in one setting.
One more lever: existing Copilot Business and Enterprise customers get temporary promotional credits for June, July, and August, $30/month for Business and $70/month for Enterprise on top of the seat allowance (GitHub blog, April 27). Plan your caps knowing that cushion expires after August.
FAQ
What happens when I hit my credit limit? Two options (GitHub Docs). Set an additional budget in US dollars and keep working at the per-token rates, where a $10 budget buys 1,000 more credits. Or wait for the next cycle, when your included allowance resets. Additional usage may be capped, and for individual plans GitHub may limit your extra credits based on usage patterns, billing history, and account verification (GitHub Changelog, June 1). If you keep hitting the wall, upgrading a tier is usually cheaper than stacking overage.
Can I still pay annually for an individual plan? Not as a new buyer. GitHub is retiring annual individual plans; current buyers pick monthly (GitHub blog, April 27). If you are already on an annual Pro or Pro+ plan, you stay on premium-request pricing until it expires, then move to Free with the option to go monthly, or you can convert early and get prorated credits for the remaining value.
Is Copilot Max available to me? Only if you already subscribe. Max opened June 1 as an upgrade for existing Student, Pro, and Pro+ subscribers, and new sign-ups across all these plans remain paused (GitHub Changelog, June 1). GitHub says sign-ups reopen in the coming weeks.
Do code completions cost credits? No. Completions and Next Edit suggestions are unlimited on every paid plan and never draw from your balance (GitHub Docs). Only the AI-model features, chat, CLI, cloud agent, Spaces, Spark, and third-party agents, spend credits.
How do I make my credits last longer? Keep auto model selection on for the 10% discount, and let it route simple work to lighter models (GitHub Docs). Model choice is the single biggest lever on spend; a frontier model can cost an order of magnitude more per token than a lightweight one for the same task.
The pick, by use case
If you mostly type and chat, Pro at $10 is the answer, and almost nobody in that bucket needs more. The moment cloud agents become a regular part of how you ship, jump to Pro+ at $39 (GitHub Docs); Pro's 1,500 credits will have you babysitting the meter. Reserve Max for the fleet-mode-plus-security-review profile that runs parallel agents all day, and remember it is upgrade-only until sign-ups reopen, so get on Pro+ first if you are not already in.
For team leads, the controls matter more than the tier: cap total per-seat usage with user-level budgets before a heavy sprint, pin a default Actions runner for code review org-wide, and bank the June-through-August promotional credits while they last. Whatever tier you land on, leave auto model selection running. The 10% discount it puts on model costs (GitHub Docs) is the easiest savings you will ever bank, and it keeps the expensive models out of the cheap work. Current pricing and the upgrade path are on the plans page.