GitHub Copilot Switches to Usage-Based Billing June 1, 2026

TL;DR

• GitHub Copilot moves to token-based billing on June 1, 2026 — no more premium request units
• Base plan prices stay the same, but you'll buy AI Credits instead of unlimited usage
• Agentic workflows now cost real money — a 10-minute chat and a 3-hour autonomous session won't be priced identically anymore
• Preview bills launch in May so you can see what you'll actually pay before the switch

The Big Picture

GitHub is ending the flat-rate era for Copilot. Starting June 1, 2026, every plan switches from premium request units to token-based billing with monthly AI Credit allotments. Your $10/month Pro plan still costs $10/month — but now it includes $10 in credits that get consumed based on actual token usage.

This isn't a surprise. Copilot has evolved from an autocomplete tool into an agentic platform that can run multi-hour coding sessions across entire repositories. The old model — where a quick question and a marathon autonomous session cost the same — was unsustainable. GitHub was absorbing massive inference costs, and the premium request cap was becoming a blunt instrument that punished heavy users without reflecting actual compute demand.

The new model aligns pricing with reality. Token consumption drives cost. Input tokens, output tokens, cached tokens — all counted at published API rates for each model. If you're running GPT-5.5 sessions or long agentic workflows, you'll burn through credits faster than someone doing basic completions. That's the point.

How It Works

Premium request units disappear. GitHub AI Credits replace them. Every plan gets a monthly credit allotment equal to its subscription price. Pro users get $10 in credits. Pro+ users get $39. Business gets $19 per seat. Enterprise gets $39 per seat.

Credits are consumed based on token usage. GitHub will charge you according to the published API rates for each model you use. Input tokens, output tokens, and cached tokens all count. The exact rates vary by model — expect frontier models like GPT-5.5 to cost more per token than older models.

Code completions and Next Edit suggestions remain free. They don't consume credits. This is critical — the bread-and-butter autocomplete experience that made Copilot popular stays unlimited. The usage-based billing only kicks in for chat, agentic workflows, and other premium features.

Fallback experiences are gone. Today, if you exhaust your premium requests, Copilot drops you to a lower-cost model and you keep working. Under the new system, when your credits run out, you stop — unless your admin has enabled additional usage or you buy more credits. This is a hard cap, not a graceful degradation.

For businesses, GitHub is introducing pooled credits. Instead of each user's unused credits sitting idle, they pool across the organization. If one developer burns through their $19 allotment and another uses $5, the org can reallocate. Admins get budget controls at the enterprise, cost center, and user levels. You can cap spend or allow overage at published rates.

One gotcha: Copilot code review will now consume both AI Credits and GitHub Actions minutes. The Actions minutes are billed at standard per-minute rates. If you're running PR reviews with Copilot, factor that into your budget.

What This Changes For Developers

If you're a light user — autocomplete, occasional chat, no agentic workflows — nothing changes. Your $10/month Pro plan still works the same way. Code completions are free. You'll probably never hit your credit limit.

If you're a power user running long agentic sessions, you'll need to budget. A multi-hour autonomous coding session that iterates across a repo will consume significantly more tokens than a quick chat question. GitHub hasn't published exact token costs yet, but expect frontier models to be expensive. If you're regularly running GPT-5.5 or Claude Opus sessions, your $10 credit allotment might not last the month.

For businesses, the pooled credits are a win. Stranded capacity was a real problem under the old model — some users maxed out their requests while others barely touched theirs. Pooling fixes that. But admins will need to monitor usage closely. The budget controls are powerful, but they require active management. If you don't set caps, a few heavy users can blow through your org's credits in days.

Annual plan holders get a grace period. If you're on an annual Pro or Pro+ plan, you stay on premium request-based pricing until your plan expires. But model multipliers increase on June 1, so your requests will cost more even before you transition. When your annual plan expires, you drop to Copilot Free unless you upgrade to a monthly paid plan. GitHub will prorate your remaining annual value if you convert early.

Try It Yourself

GitHub is launching a preview bill experience in early May. Log into github.com, navigate to your Billing Overview page, and you'll see projected costs based on your current usage patterns. This is your chance to understand what you'll actually pay before June 1.

If you're on a Business or Enterprise plan, check your admin dashboard. GitHub is automatically adding promotional credits for June, July, and August — $30/month for Business, $70/month for Enterprise. Use those months to baseline your actual usage and set realistic budgets.

For individuals, track your usage now. If you're consistently hitting premium request limits under the current model, you'll likely need to buy additional credits under the new one. GitHub hasn't published overage rates yet, but expect them to align with standard API pricing.

The Bottom Line

Use this if you're a light Copilot user who relies on autocomplete and occasional chat. Your costs won't change, and you'll benefit from the same $10/month pricing with no surprises. Skip the annual plans — they're being phased out, and the model multiplier increases make them less attractive.

Avoid this if you're a heavy agentic user on a tight budget. Long autonomous sessions will burn through credits fast, and overage costs could add up. If you're running multi-hour coding sessions regularly, budget for additional credits or consider whether the productivity gain justifies the cost.

The real risk is for businesses that don't monitor usage. Pooled credits are great, but without budget controls, a few power users can drain your org's allotment in days. Set caps early, track usage in May, and use the promotional credits to establish baselines. The flexibility is there, but it requires active management.

Source: GitHub Blog