GitHub Copilot is moving to usage-based billing

· ai ai-agents devtools · Source ↗

TLDR

  • GitHub Copilot replaces premium request units with token-based AI Credits on June 1, 2026; plan prices hold but model cost multipliers rise sharply.

Key Takeaways

  • Each paid plan includes credits equal to its monthly price: Pro gets $10, Pro+ $39, Business $19/user, Enterprise $39/user in monthly AI Credits.
  • Code completions and Next Edit Suggestions remain free and do not consume AI Credits under any plan.
  • Fallback to lower-cost models when credits run out is removed; admins get budget caps at enterprise, cost center, and user levels instead.
  • Annual plan holders stay on PRU pricing until their plan expires, then drop to Copilot Free; model multipliers increase for them on June 1 regardless.
  • Business and Enterprise customers get promotional overage credits June through August ($30 and $70 respectively) plus pooled org-wide credit balances.

Hacker News Comment Review

  • The multiplier table is the real story: Opus 4.6/4.7 goes from 3x to 27x, GPT-5.4 from 1x to 6x, Sonnet 4.6 to 9x – commenters broadly concluded OpenRouter or direct API access is now cheaper for agentic sessions.
  • The $10-for-$10-credits framing draws heavy skepticism: no bulk discount versus buying tokens directly removes the main retention argument for paying a monthly subscription over PAYG providers.
  • Emerging consensus is that Copilot’s remaining value proposition is VSCode inline autocomplete, which stays free – agentic and chat usage is better served by Claude Code or direct API access at these multipliers.

Notable Comments

  • @4ndrewl: “The Porsche you rented at $200/mo is now a Honda. But the price hasn’t changed” – sharpest summary of the plan-value erosion argument.
  • @pdp: flags Sonnet 4.6 moving from 1x to 9x multiplier and argues buying tokens directly gives more model choice at lower cost.

Original | Discuss on HN