TLDR: GitHub Copilot moved from subscription to usage based billing, and some users report up to 100 fold bill jumps.
Key Takeaways:
- Microsoft switched GitHub Copilot to usage based pricing, pushing developers to track token like credits more tightly.
- Reported complaints cite subscription plan credits of 1,500 Pro, 7,000 Pro Plus, and 20,000 Max credits, with bills spiking.
- The model choice and long running chats can burn through credits fast, creating agent risks and cost surprises.
The new pricing model turns AI coding into a budget test, not a subscription perk. For cautious developers, the cost meter still feels allergic to restraint.
The new pricing model turns AI coding into a budget test, not a subscription perk. For cautious developers, the cost meter still feels allergic to restraint.
Q&A
Why do some Copilot users claim their bills jump even with careful usage?
Token consumption can spike when prompts include long context or when conversations are re sent, so what feels small in writing can be big in tokens.
What should teams change first when they adopt usage based AI tools?
They should set internal rules for model selection, cap conversation length, and audit which tasks trigger the most token burn.
How can switching the underlying model swing costs so dramatically?
Different models price token usage differently, so the same coding request can consume far more credits depending on model choice.
What happens when AI is used as an agent rather than a helper?
Agents can run longer loops and take additional actions, so credit usage grows faster and unchecked behavior can multiply the bill.
Is there a precedent for this kind of backlash in AI software pricing?
Yes. Prior pricing changes across AI coding and chat products have triggered similar complaints, especially when allowances are hard to forecast per task.
No comments yet. Be the first to share your thoughts!