Apr 28, 2026

GitHub Copilot's Pricing Rollercoaster: A Masterclass in Losing Developer Trust

GitHub changed Copilot pricing four times in April alone. Here's the full timeline, what it means, and why Cursor is laughing. 🎢

AI & ML

Tags (5)

# github-copilot # cursor # claude-code # ai-tooling # developer-experience

Imagine paying $10/month for a gym membership. One morning you show up and the squat rack has a velvet rope around it. “Oh, that’s Pro+ now. $39/month. But here’s a nice resistance band.”

That’s GitHub Copilot in April 2026.

In the span of three weeks, GitHub paused trials, killed model access, froze signups, and announced a billing overhaul. If you blinked, you missed at least one announcement that devalued your current plan. Let me walk you through the wreckage.

The April timeline: a drama in four acts

Here’s what happened, in order, because even I had trouble keeping track.

April 10 — The trial freeze. GitHub paused new Copilot Pro trials, citing “a significant rise in abuse of our free trial system.” By April 13, existing trials got suspended too. Translation: too many people tried the free thing, so nobody gets to try anymore.

April 10 — Opus gets the axe (for Pro). Same day, GitHub announced Opus 4.6 Fast retirement from Pro+, and started enforcing stricter usage limits. The real kicker came ten days later.

April 20 — The big one. GitHub dropped the blog post titled “Changes to GitHub Copilot plans for individuals.” Buried in the corporate prose: Opus models removed entirely from Pro. Only Pro+ ($39/month) keeps Opus 4.7. Older Opus versions (4.5, 4.6) removed from Pro+ too. New signups for Pro, Pro+, and Student plans — all paused. Oh, and “stricter usage limits now apply.” Refunds available through May 20, which is corporate for “we know you’re mad.”

April 22 — Business signups paused. Even organizations on Free and Team plans can’t self-serve into Copilot Business anymore. Existing customers keep their seats, but the door is shut for everyone else.

timeline
  title GitHub Copilot Drama Timeline
  section Week 1
      Apr 10 : Pro trials paused
             : Opus 4.6 Fast retired from Pro+
             : Stricter limits enforced
  section Week 2
      Apr 13 : Existing trials suspended
      Apr 16 : Opus 4.7 released (Pro+ only)
  section Week 3
      Apr 20 : Opus removed from Pro entirely
             : All individual signups paused
             : Refund window opened
      Apr 22 : Business self-serve signups paused
  section Coming
      Jun 1  : Usage-based billing transition

The pricing shell game

Let’s look at what you actually get for your money now.

See the pattern? The $10 tier lost its most capable model. The $39 tier is the new floor for anything interesting. And if you’re new? You can’t even sign up for either.

This is the classic SaaS bait-and-switch. Launch generous, acquire users, tighten the screws once there’s lock-in. GitHub didn’t invent this playbook — but they’re executing it with impressive speed. Most companies take years to erode their value prop. GitHub did it in a single April.

And it gets better. On June 1, 2026, GitHub transitions from request-based billing to usage-based billing for both organizations and individuals. What does “usage-based” mean exactly? We don’t fully know yet. But if your premium requests today feel limited at 300/month, wait until each one has a variable cost attached to it.

Additional premium requests already cost $0.04 each. At that rate, a heavy user burning 1,000 requests/month on Pro would pay $10 + $28 in overages = $38/month. Conveniently close to Pro+ pricing. Funny how that works.

Why this happened: the economics of AI-assisted coding

Let me be fair to GitHub for one paragraph.

Running large language models is expensive. Anthropic’s Opus isn’t cheap. OpenAI’s GPT-5.x family isn’t cheap. Every premium request that hits Claude Opus costs GitHub real money. When you’re charging $10/month and users are hammering the most expensive model available, the math doesn’t work. It’s the same reason unlimited data plans disappeared from mobile carriers — “unlimited” only works when most people don’t use much.

GitHub’s problem is that developers do use much. Power users treat Copilot like a conversation partner, not a suggestion engine. They run agentic coding sessions, pipe entire codebases through context windows, and expect Opus-quality responses on every prompt. At $10/month, GitHub was subsidizing each heavy user to the tune of probably $50-100/month in API costs.

So the correction was inevitable. The execution of that correction? That’s where GitHub fumbled.

The competitive landscape: who benefits from the chaos

While GitHub was busy rearranging deck chairs, the competition was eating their lunch.

Cursor: the $2B gorilla nobody expected

Cursor’s recurring revenue doubled in three months to $2 billion. That’s not a typo. Two billion. In recurring revenue. For an AI code editor that didn’t exist three years ago.

Why? Because Cursor did the opposite of what GitHub is doing. Instead of restricting model access, Cursor leaned into model diversity. You pick your model. You pick your context strategy. The editor gets out of your way. And critically — no “premium request” anxiety. You know what you’re paying, you know what you’re getting.

Cursor also ships fast. New models show up in days, not weeks. When Anthropic drops a new Claude version, Cursor users have it before GitHub finishes writing the changelog post.

Claude Code: the terminal-native option

Anthropic’s own Claude Code takes a different approach entirely. No IDE wrapper. No monthly seat license. You bring your API key, you pay per token. It runs in your terminal, reads your codebase, and acts more like a senior engineer pair-programming with you than an autocomplete engine.

The economics are transparent: you pay exactly what you use. A light day costs pennies. A heavy refactoring session might cost a few dollars. But there’s no tier anxiety, no “you’ve hit your premium request limit,” no waking up to find your favorite model was removed overnight.

For engineers who live in the terminal — and let’s be honest, the best ones usually do — Claude Code is already the default. It doesn’t try to replace your editor. It augments your workflow.

OpenAI Codex: the new entrant

OpenAI launched Codex as an agentic coding tool — asynchronous, sandboxed, designed for larger tasks. It’s early, but it targets a different workflow: you describe a task, Codex works on it in the background, you review the result. More like delegating to a junior dev than pair-programming.

The pricing model matters here too. OpenAI charges per-task rather than per-seat, which aligns costs with value delivered. You don’t pay for the tool to sit idle.

A fair technical comparison

Let me compare these tools on what actually matters for day-to-day engineering work.

The pattern is clear. GitHub Copilot’s advantage was always distribution — it was right there in VS Code, the most popular editor. But distribution only buys loyalty until users get burned. And users are getting burned.

Predictions: where this is heading

Here’s my read on what happens next.

Short term (Q2 2026): GitHub reopens signups with the new usage-based billing. Prices will look similar on paper but cost more for power users. Opus stays Pro+ only. Some users come back out of convenience. Most who left don’t.

Medium term (H2 2026): GitHub bundles Copilot more aggressively into GitHub Enterprise. The individual developer market becomes secondary. Microsoft’s real play is enterprise seats, where switching costs are high and procurement doesn’t comparison-shop like individual devs do.

Long term (2027+): The AI coding tool market fragments into three tiers. IDE-integrated copilots (GitHub, Cursor) for everyday work. Terminal agents (Claude Code) for complex engineering. Async task runners (Codex) for batch work. Most serious engineers use at least two of these.

The winner of the individual developer market won’t be the one with the best model — models are commoditizing. It’ll be the one with the most honest pricing and the least bait-and-switch. Right now, that’s not GitHub.

The real lesson

GitHub’s April pricing saga isn’t really about pricing. It’s about trust.

Developers are a specific kind of customer. They read changelogs. They notice when features disappear. They share notes on Reddit and Hacker News within minutes. You can’t quietly remove Opus from a plan and hope nobody notices. They will notice. They will screenshot the before and after. They will write blog posts about it.

The irony is thick: GitHub — the platform built on open-source transparency — is learning that opacity in pricing destroys trust faster than any competitor ever could.

If you’re currently on Copilot Pro, grab that refund before May 20 and spend a weekend with Cursor or Claude Code. You might find that the tool you were paying for isn’t the one you need.

And if you’re on Pro+ and happy? Keep it. Opus 4.7 is genuinely good. Just don’t be surprised when the next changelog post moves your cheese again.

The squat rack will always end up behind the velvet rope. The question is whether you keep paying for the gym.