GitHub just announced that Copilot's moving to usage-based billing. Starting June 1, 2026, every plan shifts from premium request pricing to a credit system where you consume GitHub AI Credits based on token usage. Base prices stay the same: Pro at $10/month, Pro+ at $39/month, Business at $19/user/month, Enterprise at $39/user/month. But what you get for that money is about to change.
The credit system comes with multipliers that make advanced models expensive fast. According to GitHub's documentation, newer GPT and Claude Sonnet models cost 6x more credits than base models. Claude Opus costs 27x. And credits expire monthly. Use them or lose them. Hacker News commenters quickly flagged this as the end of Microsoft's subsidized inference era, where the company absorbed costs to drive adoption. That era's officially over.
The shift makes sense from GitHub's perspective. Copilot's evolved from simple autocomplete into something that runs multi-step agent sessions, burning through tokens at rates that flat pricing can't sustain. But it opens a real gap for competitors. Cursor, the VS Code fork, lets you bring your own API keys and pay providers directly. Codeium runs its own optimized models to keep costs down. Both avoid the use-it-or-lose-it credit problem.
For casual users on base models, nothing changes. Power users who've grown reliant on Claude Opus for complex refactoring and agent workflows are about to feel the actual cost of inference. Some will pay up. Others will start shopping.