As of today, GitHub Copilot has moved from flat-rate subscriptions ($10–$39/mo)
to a token-usage billing model. Developer reaction has been swift and negative,
with some heavy users reporting projected costs exceeding $2,000/month.
The new model
Instead of unlimited completions and chats, Copilot now meters every interaction:
| Usage Level |
Old Price |
New Price (est.) |
| Light (occasional completions) |
$10/mo |
~$15–25/mo |
| Moderate (daily driver) |
$10–39/mo |
~$50–150/mo |
| Heavy (agent mode, long sessions) |
$39/mo |
~$500–2,000+/mo |
Who gets hit hardest
Teams using Copilot's agent mode for multi-file refactors and long chat sessions
are seeing the steepest increases. The agent mode consumes significantly more
tokens than inline completions, and there's no cap on per-user spend.
Alternatives gaining traction
- Claude Code (Anthropic), leads in developer satisfaction (46% of devs
picked it as favorite in JetBrains' April 2026 survey), flat pricing
- Cursor, surpassed $2B ARR, used by 64% of Fortune 500 orgs
- OpenAI Codex, free for enterprise through July 14 with one-click
migration from Copilot
- Gemini CLI, Google's $100/mo developer plan includes coding agents
The price war context
This pricing change lands in the middle of an AI coding tools price war.
OpenAI made Codex free for enterprise for two months starting May 14, and
Anthropic increased Claude Code limits by 50% through July 13. Both are
burning cash to lock in users before IPOs later this year.
The bottom line
If your team is on Copilot, check your projected bill now. The alternatives are
mature and many are cheaper. If you're a heavy user, the math may push you
toward Claude Code or Cursor sooner than you planned.