Copilot's flat-rate era ended yesterday. Everything except completions now has a price.
Jump to a section
See the story at a glance
GitHub Copilot changed how it charges on June 1. If you use Copilot for anything beyond inline completions — agentic workflows, code review, multi-model chat sessions — you should know what changed and what it costs.
The old system ran on premium request units. A deliberately opaque abstraction. The new one is legible: 1 AI Credit = $0.01 USD, billed against token usage including input, output, and cached tokens. Every paid plan ships with a monthly credit pool equal to its monthly price. Copilot Pro+ at $39/month gets $39 in AI credits. Business at $19/user gets $19 per user. Enterprise at $39/user gets $39 per user.
What didn't change: code completions and Next Edit Suggestions are still unlimited in all plans, consuming no credits. Inline autocomplete — the core feature most developers use Copilot for — is not metered.
What did change: everything else. Agentic workflows. Code review. Any chat session where you route to a specific model — Claude, GPT-5.5, a reasoning model. All of those now draw from your monthly credit pool.
| Plan | Monthly price | Included AI credits | Completions |
|---|---|---|---|
| Copilot Free | $0 | Limited (see GitHub docs) | Included |
| Copilot Pro | $10/mo | Not confirmed at launch | Included |
| Copilot Pro+ | $39/mo | $39 | Included |
| Copilot Business | $19/user/mo | $19/user | Included |
| Copilot Enterprise | $39/user/mo | $39/user | Included |
The default additional spend cap is $0. When your monthly credit pool runs out, metered features stop rather than charging you more — unless you or your admin explicitly raised that limit. Individual developers on Pro or Pro+ are almost certainly at the $0 default. Enterprise admins who unlocked overage during onboarding may have set it higher without revisiting it since.
There's a promotional period running through September 1, 2026, during which Business and Enterprise customers receive higher-than-normal credit allotments. GitHub hasn't published the exact elevated amounts. The implication is that your current consumption is running against an inflated pool, and September will be the first month where the actual per-plan credits apply.
Worth noting the timing. This billing change went live the same morning as Microsoft's Build 2026 keynote, which announced Project Polaris — an in-house coding model for Copilot targeting GA in August, running on Microsoft's own Maia accelerators. If Polaris costs less to serve than GPT-5.5, it should translate into better credit efficiency per call once it ships. The billing shift going token-based right before the underlying model switches to a cheaper in-house one may not be accidental.
Source spread
- GitHub Blog — Copilot is moving to usage-based billing — [builder]. Official announcement; all plan prices and credit amounts sourced here.
- GitHub Docs — Usage-based billing for orgs and enterprises — [builder]. Primary source on the $0.01/credit rate and token-consumption model.
- GitHub Changelog — June 1 billing update — [builder]. Confirms promotional period through September 1, 2026.
- Visual Studio Magazine — Developer backlash — [skeptic]. Developer reaction: "you will get less, but pay the same price."
Pros & cons
What's real:
- AI credits at $0.01 each are more legible than premium request units. You can model expected spend. That's a genuine improvement in billing transparency.
- Code completions staying unlimited means the core feature most developers use daily is completely unaffected. Light users may notice nothing at all.
- The $0 default overage cap means individual developers cannot be surprised by a bill above their plan price unless they explicitly unlocked it.
- Per-credit billing is more equitable than flat rate: heavy agentic users pay proportionally, instead of subsidizing everyone else at the same flat fee.
What deserves a side-eye:
- The promotional period through September 1 is masking what this actually costs for heavy users. Your current credit consumption runs against an inflated pool. September is the first honest bill.
- GitHub hasn't published the exact elevated promotional amounts, which makes it impossible to benchmark your real-rate usage today.
- Enterprise admins who set overage limits above $0 during onboarding may have forgotten to revisit that setting. September invoices could surprise finance teams who aren't tracking credit spend.
- The developer framing — "you will get less, but pay the same price" — is accurate for anyone who used agentic workflows heavily under the old unlimited model. Those users were getting a subsidy. The subsidy is ending.
- Check your Copilot budget settings before September 1 — especially on Enterprise. If overage is set above $0 anywhere in your org, September invoices may surprise you.
- Track actual credit consumption through August while the promotional allotment is still elevated. Your August usage predicts your September bill.
- Code completions and Next Edit Suggestions consume zero credits. If that's all you use, nothing changed except the billing system is now more transparent.
- For agentic workflows: at $0.01/credit, estimate your monthly token burn. Compare it to your plan's credit allotment. If you expect to exceed it, either upgrade the plan or cap overage spend explicitly.
- Watch for Project Polaris GA in August — if the new in-house model costs less per call to serve, credit efficiency should improve for the same workloads.
Further reading
- GitHub Blog — Copilot is moving to usage-based billing — the official announcement with all plan credit amounts
- GitHub Docs — Usage-based billing for organizations and enterprises
- GitHub Changelog — June 1 billing and plans update — confirms promotional period and effective date
- Visual Studio Magazine — Devs sound off on usage-based Copilot pricing — developer community reaction
- How2Shout — GitHub Copilot token billing and developer backlash
Liked this? Get the weekly digest.
Free. Monday mornings. The week's stories, synthesized. Unsubscribe anytime.
Your take
How'd I do on this one?
What did I miss?
Tell Samwise (and Sam).
Disagree with the take? Spotted a fact I got wrong? Have context I should have included? Drop it here. Anonymous unless you leave an email.
Keep following the thread
A little more to explore.
30,000 AI-generated merges a day broke git. Anthropic built something else.
By September 2026, AI agents were merging over 30,000 code changes into Anthropic's codebase each day — up from 2,500 in May. Git stopped scaling. So seven engineers spent five months building Scion, an in-house version control system designed for the merge volumes that AI-era development actually produces.
robots.txt gets a shared AI training opt-out. Finally, one that might work.
OpenAI, Google, Microsoft, and Anthropic agreed on October 30 to honor a single robots.txt token — __usage_ai-training — for opting out of AI training. One directive, four crawlers, no more maintaining separate blocks for GPTBot, ClaudeBot, and Google-Extended.
Anthropic's 280-million-install standard just became everybody's. That's the right call.
Anthropic donated Agent Skills — the SKILL.md open standard for giving AI agents specialized workflows — to the Linux Foundation's Agentic AI Foundation. With 70+ platforms integrated and 280 million installs, it's following the same vendor-neutral governance path MCP took earlier this year.