In late April 2026, GitHub quietly updated its Copilot billing policy: starting June 1, all Copilot Pro and Pro+ annual subscribers will switch from per-request to per-token billing, with significant multiplier increases for frontier models.
This signals the end of the era of cheap coding tool subsidies.
Multiplier Changes at a Glance
According to the updated GitHub documentation, the model multiplier adjustments are substantial:
| Model | Old Multiplier | New Multiplier | Increase |
|---|---|---|---|
| Claude Opus 4.6 | 3x | 27x | 900% |
| Claude Opus 4.5 | 3x | 15x | 500% |
| Claude Opus 4.7 | 7.5x | 27x | 360% |
| Claude Sonnet 4.6 | 1x | 9x | 900% |
| Claude Sonnet 4.5 | 1x | 6x | 600% |
| Gemini 3 Pro | — | 6x | New |
| Gemini 3.1 Pro | — | 6x | New |
| GPT-5.1 | — | 9x | New |
| GPT-5.5 | — | 27x | New |
For a Copilot Pro subscriber at $19/month, under the per-request model, using Opus 4.6 daily consumed roughly 5% of the usage quota. Under the new per-token billing, the same usage level will cost 9 times more. The Reddit ClaudeAI community has seen significant backlash, with some users stating they will not renew after June.
Context: The Subsidy Contraction
GitHub Copilot’s multiplier adjustment is not an isolated event. In the same time window, Anthropic quietly changed Opus model access for Claude Pro users—Pro users must now enable API billing to use Opus models in Claude Code. This is effectively a “paywall within a paywall.”
Together, these moves point to a clear trend: from 2025 to early 2026, model companies heavily subsidized AI coding tools to capture developer market share. Now, with the market格局 largely set (Claude Code dominates, OpenAI Codex follows), the subsidies are being withdrawn.
Opportunity for Open Source Alternatives
After Anthropic removed Claude Code from the $20 Pro plan, the developer community responded quickly. Open source terminal agent projects like OpenClaude gained significant attention—supporting any model (GPT, Gemini, local), no subscription limits, running full coding agent workflows in the terminal.
When the cost of closed-source tools begins to approach API prices, “bring your own model” open source solutions become more attractive to heavy users.
Market Outlook
GitHub Copilot’s multiplier adjustment marks the transition of AI coding tools from the “customer acquisition subsidy phase” to the “commercialization recovery phase.” For developers, this means reassessing model choices after June: the cost of high-frequency Opus-level model usage will rise significantly, while Sonnet-level and open source models (like Kimi K2.6) become relatively more cost-effective.
For model companies, this is a rebalancing between market share and profit margins. Whoever can maintain user retention after price adjustments truly owns pricing power.
Sources
- GitHub Copilot Billing Documentation
- Reddit ClaudeAI Community Discussion
- X platform developer analysis (@mjovanovictech and others)