Skip to content

GitHub Copilot Complete Guide

The Real Problem Behind GitHub Copilot's Individual Plan Tightening — The Request-Based Billing Model Has Hit Its Limit

For / Key Points

For: GitHub Copilot individual plan (Pro / Pro+) users, and developers or product managers tracking pricing model trends in AI coding tools.

Key Points:

  • The trigger was a bug fix — After March's token undercounting bug was fixed, rate limit incidents surged across the community1.
  • The structure is a pricing model breakdown — Request-based billing can no longer distribute costs across next-gen model workloads2.
  • Token-based billing is the destination — The current restrictions are positioned as pre-migration cleanup before a shift to token billing2.

What Changed

On April 21, 2026, GitHub announced three changes to Copilot individual plans3: a temporary halt to new signups, stricter usage limits, and adjustments to Opus model availability. Timing and scope of each change are summarized below.

ChangeEffective DateScope
New signup haltApril 20, 20264Pro / Pro+ / Student
Usage display in VS Code / CLIStaged rollout from April 213All plans
Stricter usage limitsPhased over "coming weeks"3Pro / Pro+
Opus 4.7 promo multiplier (7.5×)Until April 30, 20265Pro+ only
Removal of Opus 4.5 / 4.6 from Pro+Coming weeks6Pro+ only
Refund request windowApril 20 – May 20, 20263Pro / Pro+

The most significant detail: Opus-tier models are disappearing from the Pro plan entirely. Pro+ (39/month) will retain only Opus 4.7 at the 7.5× multiplier, while Opus 4.6 (3×) and 4.5 are being phased out[^6]. Pro (10/month) users who want Opus-tier output now face a nearly 4× price jump to Pro+ as their only option.

March's Bug Fix as the Trigger

What this section answers: why April, and why now?

In March 2026, GitHub identified a bug in Copilot's rate limiting system that caused tokens to be undercounted1. The affected models — Claude Opus 4.6, GPT-5.4, and other next-gen releases — consume significantly more infrastructure per request than older models, but the bug was measuring them as artificially lightweight1.

After the fix, the situation changed immediately. With real costs now reflected accurately, community reports emerged of Pro+ users hitting weekly rate limits after consuming only around 20% of their nominal request quota7. As The Register reported, some community members framed it less as a bug fix and more as a business decision1:

This feels more like a business decision than a bug fix — subsidization had reached an unsustainable level1

The causal chain from that point to the April 21 announcement is straightforward: token undercounting discovered → bug fixed → rate limit incidents spike → new signups halted → usage limits tightened.

The Structural Flaw in Request-Based Billing

What this section answers: why were tighter limits the only option?

Copilot's billing unit is the "premium request." Pro includes 300/month; Pro+ includes 1,500/month, with overages at $0.04/request8. Each model carries a multiplier — heavier models consume more requests per invocation5.

The premise of this design was that plan pricing could absorb the average cost per request within a manageable range. Next-gen models broke that premise. Opus 4.7 carries the same API unit price as Opus 4.6 — 5/25 per million tokens (input/output) — but its Copilot multiplier jumped from 3× to 7.5×, a 2.5× increase9. The same underlying cost now carries 2.5× the "price" in request units, exposing how far the abstraction has drifted from actual costs.

ModelAPI price (in/out)Copilot multiplierPro+ monthly quota
Opus 4.6$5 / $253×500 uses
Opus 4.7 (promo)$5 / $257.5×200 uses
Opus 4.6 Fast (removed)$5 / $2530×250 uses

Identical underlying cost, yet the "value of one request" spans a 6× range across three models. Internal documents reported by Where's Your Ed At indicate that weekly Copilot operating costs roughly doubled from the start of the year — signaling a breakdown in overall unit economics, not just an isolated model issue2.

Token-Based Billing Is Now the Default Trajectory

What this section answers: where does this lead?

Leaked internal documents indicate GitHub is moving toward "token-based billing" — metered charges tied to actual token consumption2. The current signup halt and usage restrictions are positioned as pre-migration cleanup ahead of a pricing model overhaul from request units to token units. The transition is phased because GitHub needs to manage continuity for existing users while bringing operating costs under control.

The shift to token billing has the potential to fundamentally change what an individual plan means. The core user experience of fixed-price subscriptions — "I know my monthly ceiling" — is difficult to preserve under token metering. Competing tools like Anthropic's direct API, Cursor, and Claude Code already offer a mix of "monthly base + overage metering" and "fully metered via API key" models, and Copilot appears to be converging on this spectrum.

The underlying driver is structural across all AI coding tools. As agentic workflows have matured — with dozens of model calls per task becoming routine — the assumption that "1 request = 1 human intent" has collapsed3. GitHub's blog post acknowledging that "a small number of requests can now cost more than the plan price" is an official admission of that collapse3.

Practical Impact for Working Developers

What this section answers: what should you do right now?

For working developers, the sharpest operational pain isn't Opus 4.7's multiplier or the Opus removal from Pro. It's the undisclosed weekly token limit — a cap that can halt a session regardless of how much premium request quota remains10. Even on Pro+, only a relative figure ("more than 5× Pro") is published; the absolute number remains private3.

Practical responses available today:

  • Prioritize Auto mode: When using Copilot CLI in auto mode, all paid subscribers receive a 10% multiplier discount (a 1× model counts as 0.9×)11.
  • Suppress parallel workflows: Commands like /fleet that run in parallel double token consumption. Avoid these when approaching limits3.
  • Insert plan mode: Using plan mode before implementation both improves task success rates and reduces token consumption3.
  • Scope with AGENTS.md: Explicitly specifying files and directories to avoid full-repo scans is an established practice among Pro+ users for reducing token consumption7.

Starting April 21, 2026, VS Code and Copilot CLI display warnings at 75% and 90% of weekly quota3. This is a quiet but meaningful UX improvement aimed at eliminating the "session stops without warning" experience — and actively monitoring these alerts can substantially reduce unplanned limit hits.

Summary

This announcement is not a pricing adjustment for a single service. It is the opening act of a pricing model redesign that the entire AI coding tools industry is being forced to undertake in the agentic era. Now that the simple abstraction of flat-rate × request can no longer absorb next-gen model costs, some form of token-level consumption metering entering the pricing structure is unavoidable.

The key variable is whether individual developers interpret this change as a "price increase" or as "making real costs visible." The first framing makes migration to other services rational; the second makes token-efficient prompting and tool selection a durable long-term skill. Whichever stance you take, it's worth recognizing that this change is not an isolated event — it's the starting point of a long transition.

Post this article Copy link

  1. The Register, "Customers revolt as GitHub Copilot 'fixes' rate limits" (April 15, 2026): https://www.theregister.com/2026/04/15/github_copilot_rate_limiting_bug/ ↩↩↩↩↩

  2. Where's Your Ed At, "Exclusive: Microsoft To Shift GitHub Copilot Users To Token-Based Billing, Tighten Rate Limits" (April 2026): https://www.wheresyoured.at/news-microsoft-to-shift-github-copilot-users-to-token-based-billing-reduce-rate-limits-2/ ↩↩↩↩↩

  3. GitHub Blog, "Changes to GitHub Copilot Individual plans" (April 21, 2026): https://github.blog/news-insights/company-news/changes-to-github-copilot-individual-plans/ ↩↩↩↩↩↩↩↩↩↩

  4. GitHub Docs, "About individual GitHub Copilot plans and benefits": https://docs.github.com/en/copilot/concepts/billing/individual-plans ↩

  5. GitHub Docs, "Supported AI models in GitHub Copilot": https://docs.github.com/en/copilot/reference/ai-models/supported-models ↩↩

  6. GitHub Changelog, "Claude Opus 4.7 is generally available" (April 16, 2026): https://github.blog/changelog/2026-04-16-claude-opus-4-7-is-generally-available/ ↩

  7. GitHub Community Discussion #192485, "GitHub Copilot Pro+ rate limited for hours": https://github.com/orgs/community/discussions/192485 ↩↩

  8. GitHub Docs, "Requests in GitHub Copilot": https://docs.github.com/en/copilot/concepts/billing/copilot-requests ↩

  9. GitHub Community Discussion #192814, "GitHub Copilot Claude Opus 4.7 pricing not correct": https://github.com/orgs/community/discussions/192814 ↩

  10. GitHub Docs, "Rate limits for GitHub Copilot": https://docs.github.com/en/copilot/concepts/rate-limits ↩

  11. GitHub Changelog, "GitHub Copilot CLI now supports Copilot auto model selection" (April 17, 2026): https://github.blog/changelog/2026-04-17-github-copilot-cli-now-supports-copilot-auto-model-selection/ ↩