The $10 Copilot Era Is Over: GitHub Switches to Token Billing on June 1

Crucially, code completions and Next Edit Suggestions remain free and unmetered. The billing applies to higher-compute features like chat, agentic sessions, CLI calls, and code reviews . Once you exhaust your monthly credit pool, Copilot stops responding—there is no degraded mode or silent downgrade unless you set a budget for additional usage .

Another notable change: Opus models have been removed from the Pro plan and are now only available to Pro+ subscribers . New sign-ups for Pro, Pro+, and Student plans were also paused starting April 20, 2026, ahead of the transition .

Promotional credits soften the initial blow

To ease the transition, GitHub is offering promotional credit bumps for existing customers. Internal documents revealed that from June through August 2026, Business customers paying $19/user/month will receive $30 of pooled AI Credits, while Enterprise customers at $39/user/month will receive $70 of pooled credits . After the promotional period, credit allotments drop to match the base subscription price exactly . The exact promotional amounts for individual Pro and Pro+ plans have been less clearly detailed in GitHub's public documentation.

Developer reaction: from frustration to fury

The announcement triggered immediate and widespread backlash across Reddit, X, and developer forums. The dominant sentiment is that GitHub lured users with a simple $10 flat rate and is now switching to a metered model once dependency is established .

• Reported cost spikes: Some developers claim their costs could jump from $29/month to $750, and one report suggested a worst-case spike from $50 to $3,000 under heavy agentic usage .
• "What a joke": TechCrunch captured the mood with a headline quoting a Redditor who saw their bill ballooning nearly 25x overnight .
• Threats to switch: Developers are openly evaluating alternatives like Cursor, Claude Code, Amazon Q Developer, and open-source options .
• Opus removal compounds anger: Losing access to Opus models on the base Pro tier adds insult to injury for users who now face both restricted model access and higher metered costs .

GitHub has positioned the change as offering more granular control and potential savings for light users, but the developer community views it as a significant price increase masquerading as usage flexibility . The loss of predictable flat-rate certainty is the central pain point.

The bigger picture: all AI coding tools are going metered

GitHub's move does not happen in isolation. By mid-2026, every major AI coding assistant has adopted metered, token-based usage models :

• Cursor switched from fixed requests to a $20/month credit pool, with Pro+ at $60 and Ultra at $200. Cursor's own documentation notes that daily agent users typically need $60–$100/month in total usage .
• Claude Code (Anthropic) uses a $20/month Pro tier with token-based overage, and heavy users on Max bundles can reach $100–$200/month .
• OpenAI Codex moved to token-based pricing in April 2026 .
• Amazon Q Developer charges $19/user/month with usage limits .

Industry analysis now pegs the real cost of heavy agentic AI coding tool usage at $200–$600/month per developer once overage and multiple tools are factored in . While GitHub's $10/month Pro tier remains the cheapest headline price in the market, analysts and power users agree the effective cost under the new model will be far higher. The $20/month tier—occupied by Cursor and Claude Code—has become the new baseline for serious AI-assisted development .

The message from the market is clear: the all-you-can-eat AI coding buffet is closed. The question for developers is no longer whether they will pay metered rates, but how much.