GitHub Copilot Switches to Token-Based AI Credits Billing — Effective June 1
GitHub replaced its Premium Request Units system with AI Credits, aligning billing to actual token usage per model. Code completions remain free; plan prices are unchanged, but agentic workloads could bust monthly quotas.
As of June 1, 2026, GitHub Copilot no longer charges by Premium Request Units. It charges by AI Credits — a token-based currency tied directly to model API rates. Every plan includes a monthly credit bundle equal to its subscription cost. Plan prices are unchanged.
Here’s what that means in practice: Copilot Pro ($10/month) gets $10 in AI Credits. Copilot Pro+ ($39/month) gets $39. Copilot Business ($19/user/month) gets $19 plus a promotional top-up of $30 extra credits through August 2026. Copilot Enterprise ($39/user/month) gets $39 plus $70 extra through August. Code completions and Next Edit Suggestions stay free on every plan — no credits consumed.
The shift from PRUs to token rates is a meaningful change in transparency, and also risk. PRUs were a flat abstraction — you got N requests regardless of how expensive the underlying model call was. AI Credits expose the actual cost stack: GPT-4.1, Claude Sonnet, Gemini 1.5 Pro, and GitHub’s own models each have different credit consumption rates per 1,000 tokens. Heavy agentic workflows — Copilot Workspace multi-step tasks, multi-file edits, long-context operations — can consume credits at rates that PRUs masked.
GitHub’s official position is that for typical usage patterns, the included credit bundle covers a month of work. The June-August promotional credits for Business and Enterprise plans suggest GitHub expects some users to blow past the baseline during the transition period as they adjust usage habits.
This is where developer backlash is legitimate. Agentic coding is explicitly the direction GitHub has been pushing: autonomous pull request generation, multi-step code review, cross-repo changes. Those workflows are token-heavy by design. Billing changes that make those use cases suddenly opaque or expensive cut against the pitch GitHub has been making.
On the upside: GitHub is shipping usage dashboards so teams can see credit consumption per user, per model, and per feature before overages hit. Organizations can set hard limits or soft alerts. That visibility was absent under PRUs.
For individual developers: if your Copilot usage is primarily completions and chat, you won’t notice a difference. If you’re running Copilot Workspace tasks daily or using multi-agent features, check the dashboard once it goes live and benchmark your actual consumption against the included credits before assuming the monthly bundle is enough.
The pricing docs are updated at github.com/features/copilot with per-model credit rates.