⚡ Quick Answer: GitHub confirmed on April 28, 2026 that all Copilot plans switch to usage-based billing on June 1. Copilot Pro stays at $10 — but $10 now buys 1,000 AI Credits. A developer asking 10 Copilot Chat questions daily with Claude Sonnet 4.6 burns ~60 credits per day and runs out by day 17. Tab completions remain unlimited and are completely exempt from the credit system on all paid plans. Annual subscribers can cancel for a prorated refund before May 20, 2026. That window is closing.
GitHub says Copilot Pro is still $10/month. Technically, that's true. Practically, what $10 buys is changing.
What they buried in the announcement: starting June 1, that $10 buys a fixed pool of AI Credits. Burn through it — and plenty of daily chat users will — and you're cut off for the rest of the month. Same $10. Different product. Flat-rate users aren't losing Copilot — they're losing predictability.
The developer community's reaction inside GitHub's own discussion threads has been blunt. Phrases like 'exodus gonna be epic' and 'kick in the nuts' are not edge cases — they're the top-voted comments. And yet most developers haven't actually sat down and done the math on whether their personal usage survives June 1. This article does it for you.
What Changed — and the Three Details GitHub Didn't Lead With
GitHub's Chief Product Officer confirmed in an April 28 blog post that Copilot is moving from premium request unit billing — where a quick question and a six-hour agentic session cost the same number of requests — to token-based billing at published API rates. The reason is straightforward: Copilot's weekly infrastructure cost has nearly doubled since January 2026. The flat-rate subsidy is finished.
- The Opus 4.7 multiplier is jumping 3.6x for annual plan holders. Claude Opus 4.7 — the model most developers are excited to use for agentic work — goes from a 7.5x request multiplier to 27x on June 1 for annual subscribers. One serious autonomous coding session against Opus 4.7 can now consume the entire $10 Pro allotment in a single run. This was not in the headline. It was buried in a Register article sourcing internal documents.
- Code review now also burns GitHub Actions minutes. Copilot code review moved to an agentic architecture and starting June 1, reviewing a pull request with Copilot counts against your Actions minutes at standard per-minute rates on top of AI Credits. Developers who auto-review every PR should calculate both bills.
- Annual subscribers have a documented refund exit — closing May 20. If you're on an annual Copilot plan and your heavy usage was in Chat, not autocomplete, GitHub is offering a prorated refund for remaining months if you cancel before May 20. On a $100/year Pro annual plan with eight months left, that is roughly $67 back. The detail is in GitHub's billing docs. Most coverage skipped it.
Every Option — USD, GBP, AUD
| Plan / Tool | USD / month | GBP / month (approx.) | AUD / month (approx.) |
|---|---|---|---|
| Copilot Free | $0 | £0 | A$0 |
| Copilot Pro | $10 | £8 | A$16 |
| Copilot Pro+ | $39 | £31 | A$63 |
| Copilot Business | $19/seat | £15/seat | A$31/seat |
| Cursor Pro | $20 | £16 | A$32 |
| Cursor Pro+ | $60 | £48 | A$97 |
| LumiChats Monthly | ~$14 | ~£11 | ~A$23 |
| Claude Code (Claude Pro) | $20 | £16 | A$32 |
GBP and AUD are approximate conversions at April 2026 rates. Copilot pricing confirmed from GitHub's official pricing page, April 30, 2026. Cursor pricing from cursor.com/pricing, April 2026. LumiChats monthly plan at ₹1,199/month converted at April 2026 exchange rate.
The Real Math — What $10 Actually Buys After June 1
Copilot Pro's 1,000 AI Credits sounds generous until you price out a typical chat session. Claude Sonnet 4.6 — the default model on most plans — costs $3 per million input tokens and $15 per million output tokens at GitHub's published API rates. A moderate question loads 10,000 tokens of context and generates 2,000 tokens of output. That costs 6 credits. One question. Six credits.
The number that ends the debate: the $10 Pro plan's 1,000 credits covers roughly 166 moderate questions per month — 5 to 6 per day for a full 30 days. Ask 10 per day and you run out mid-month. The remaining 13 days, you fall back to free-tier access. That is the product you are actually buying after June 1. Your exact burn rate depends on context size, model choice, and workflow — GitHub's May preview bill is the real source of truth.
| Questions / Day | Credits Burned / Day | Credits Exhausted By | Verdict on $10 Pro |
|---|---|---|---|
| 3 questions | 3 | ~18 credits | Never — full month covered |
| 5 questions | 5 | ~30 credits | Never — full month covered |
| 10 questions | 10 | ~60 credits | Mid-month (day ~17) |
| 20 questions | 20 | ~120 credits | Week one ends it (day ~8) |
That last row is the most important nuance in this story — and the one most articles published this week missed. Tab completions and Next Edit suggestions do not consume a single AI Credit on any paid Copilot plan. They remain unlimited after June 1. If your Copilot workflow is primarily autocomplete with occasional chat, June 1 may change nothing for you. The developers getting burned are the ones running Copilot Chat as a daily conversational coding partner.
Who Gets Burned — Four Specific Profiles
- Freelance developers using Copilot Chat as a real-time rubber duck — debugging sessions, architecture questions, explaining error messages line by line. If you open the chat panel more than eight times per working day, the math puts you in the moderate-to-heavy bracket. The $10 plan does not cover that workflow for a full month. Upgrade to Pro+ ($39) or move to a tool that handles chat differently.
- Developers running Claude Opus 4.7 for agentic tasks inside Copilot. The multiplier jump from 7.5x to 27x means one multi-file autonomous session — reading a codebase, finding a bug, writing a fix, running tests — can eat 200 to 500 credits in a single run. The $10 plan's 1,000 credits does not survive two of those sessions per week.
- Annual plan subscribers with moderate usage who stayed on for the savings. If you signed up annually to lock in the price but are not heavily using Chat every day, you are paying annual pricing for a product whose value proposition just shifted. The refund window is open right now. It closes May 20. Cancel, collect the prorated refund, and re-evaluate on a monthly plan in June once you have real usage data from the preview bill.
- Developers double-subscribed to Copilot and a direct Claude or GPT plan. If you already pay for Claude Pro ($20) and use Copilot separately just to pipe Claude into your editor, you are paying for the same underlying model twice — once through Anthropic and once through GitHub's pass-through at API rates. That redundancy made sense when Copilot was flat-rate. It does not make sense when Copilot chat charges by the token.
Who Should Not Panic
- Developers using Copilot primarily for tab autocomplete who open chat fewer than five times a day. Five questions per day burns 30 credits — 900 credits across a full 30-day month. The $10 Pro plan covers that with 100 credits to spare. Tab completions are unlimited. June 1 changes nothing for this group.
- Backend engineers in Visual Studio writing C# or F#. Claude Code is terminal-based. Cursor runs on a VS Code fork. Neither replicates Copilot's deep PR and code-review workflow inside Visual Studio specifically. If that integration is a real dependency, upgrade to Pro+ rather than migrate and lose it.
- Enterprise and Business teams with the transitional credit cushion. GitHub is giving Business and Enterprise seats elevated allotments from June 1 through September 1, 2026. Use those five months to run a real usage audit via the preview bill before making any platform decisions mid-sprint.
Three Alternatives Worth Considering
If you have concluded the $10 Pro plan no longer fits, these are the cleanest moves. If your primary need is chat-only rather than IDE integration, there is a fourth option at the end — but read the first two first.
- Cursor Pro ($20/month). The most direct Copilot replacement for developers who live in a VS Code environment. Unlimited tab completions, unlimited Auto mode where Cursor routes to the best model without drawing from your credit pool, and a $20 monthly credit pool for manually selected frontier models. The structural advantage over Copilot's new billing: if you stay on Auto mode, your monthly cost stays flat at $20 regardless of chat volume. That predictability is exactly what Copilot Pro just lost.
- Claude Code via Claude Pro ($20/month). The right pick for developers who want the strongest model for autonomous agentic coding runs and are comfortable in the terminal. Claude Code leads CursorBench at 70% and holds the highest developer satisfaction score in the January 2026 JetBrains survey at 91% CSAT. Pair it with any free code editor for autocomplete. The combined cost is $20/month for world-class agentic coding — versus $39 for Copilot Pro+ that now runs on metered credits.
- If your main need is chat-only: LumiChats (~$14/month). Full disclosure: LumiChats is our platform. We are listing it here because it is directly relevant — if your specific frustration is paying GitHub a per-token pass-through to talk to Claude Sonnet 4.6, and your core need is AI chat for debugging, architecture questions, and code explanation rather than IDE autocomplete, there is a cheaper path. LumiChats gives you Claude Sonnet 4.6, GPT-5.2, Gemini 2.5, Grok 4.1 Fast, and an auto-routing mode from one subscription at approximately $14/month. It does not replace Copilot's IDE autocomplete or GitHub PR integrations. Copilot pricing June 2026 makes the comparison straightforward: chat-only on LumiChats costs less than half of Copilot Pro+.
The May 20 Refund Window — Act Before It Closes
GitHub is offering annual Copilot subscribers a clean exit: cancel before May 20, 2026 and receive a prorated refund for remaining months. This is not standard policy — most annual SaaS subscriptions return nothing. GitHub is returning actual money. The Copilot Pro annual plan costs $100/year. Cancel in early May with eight months remaining and you walk away with approximately $67. The Copilot Pro+ annual at $468/year with eight months left: approximately $312 back. Go to Billing settings, find the GitHub Copilot section, and select 'Cancel and refund subscription.' Monthly plan holders receive no refund — the window applies to annual plans only.
GitHub is launching a preview bill experience in early May — visible in your billing overview — that shows projected June 1 costs based on your actual usage patterns. Check it before making any decision. The preview bill is the single fastest signal on whether your current plan survives the switch. Source: GitHub Blog, April 28, 2026.
Bottom Line
Is GitHub Copilot still worth it after the June 2026 pricing change? Yes — but only if your usage fits the new math. Copilot vs Cursor comes down to one question: do you need GitHub PR integrations, or do you need predictable billing? If you use Copilot Chat more than five times a day, check the May preview bill immediately. If it confirms heavy usage, upgrade to Pro+ or move to Cursor ($20 flat with unlimited Auto mode). If you are on an annual plan you are not using heavily, cancel before May 20 for the prorated refund — that window does not reopen.
01Is GitHub Copilot still worth it after the June 2026 pricing change?
Depends entirely on how you use it. Autocomplete-only users are completely unaffected — tab completions stay unlimited on all paid plans. Copilot Pro ($10) covers roughly 5–6 moderate chat questions per day for a full month. Push past that and the value collapses fast. Pro+ ($39), Cursor ($20), or a chat-only tool makes more sense depending on whether you need GitHub IDE integration or primarily want multi-model AI access.
02Should I cancel GitHub Copilot and switch to Cursor?
Switch if you use Copilot Chat heavily and want predictable monthly costs. Cursor Pro's unlimited Auto mode keeps your bill flat at $20 regardless of chat volume — a structural advantage over Copilot's new per-token billing. Stay on Copilot if you primarily use autocomplete or need deep GitHub PR and code-review integrations that Cursor does not replicate.
03How many Copilot chat questions does the $10 plan cover per month?
Roughly 166 moderate questions — about 5 to 6 per day for 30 days. A moderate question uses 10,000 input tokens plus 2,000 output tokens with Claude Sonnet 4.6 at $3 input and $15 output per million tokens, costing 6 AI Credits. Ask 10 questions per day and the 1,000-credit allotment runs out well before month-end. Source: GitHub API rates, April 2026.
04Can I get a refund if I cancel GitHub Copilot annual plan?
Yes — if you cancel your GitHub Copilot annual plan before May 20, 2026. GitHub is offering annual subscribers a prorated refund for remaining months. Go to Billing settings, find the GitHub Copilot section, and select Cancel and refund subscription. The May 20 window applies to annual plans only. Monthly plan holders receive no refund — cancellation takes effect at end of the current billing cycle. Source: GitHub Docs, April 30, 2026.
05Is GitHub Copilot Pro+ worth $39 a month?
For daily Copilot Chat users staying in the GitHub ecosystem, yes — 3,900 credits covers roughly 21 moderate questions per day for a full month. For developers whose primary need is AI chat rather than GitHub-native integrations, Cursor Pro ($20 flat with unlimited Auto mode) or LumiChats (~$14/month for multi-model access) deliver comparable chat capability at a meaningfully lower monthly cost.
