GitHub Copilot Usage-Based Billing: What Changed and What You Need to Know
Table of Contents
- Introduction
- Why GitHub Made This Change
- What’s Changing: Premium Requests to AI Credits
- Plan Pricing: What Stays the Same
- Understanding AI Credits and Token Costs
- Model Pricing Breakdown
- What Individual Users Need to Know
- What Teams and Businesses Need to Know
- Practical Cost Examples
- How to Estimate Your Usage
- FAQ
Introduction
On June 1, 2026, GitHub Copilot switched from a premium request-based billing model to usage-based billing. This is one of the most significant changes to Copilot pricing since the product launched. The shift reflects how Copilot has evolved from a simple code completion assistant into an agentic platform capable of multi-step coding sessions and complex interactions.
If you use Copilot individually or manage it for your team, understanding the new billing model is now critical. The good news is that plan prices are not changing. The important change is how usage is measured and billed.
This article explains the new billing system, helps you estimate costs, and shares what changes for different user types.
Why GitHub Made This Change
GitHub’s announcement highlights a fundamental shift: Copilot usage patterns have changed dramatically.
The old premium request model treated a quick chat question the same as a multi-hour autonomous coding session. Both consumed the same unit of quota. As teams started using Copilot for longer, more complex tasks—especially agentic workflows where Copilot iterates across entire repositories—the hidden cost to GitHub grew without users being charged fairly.
Behind the scenes, GitHub absorbed much of this cost. But at scale, the model became unsustainable. Usage-based billing aligns pricing with actual compute and inference costs, which means:
- Heavy agentic users pay more (because they use more compute resources)
- Light users pay less
- Pricing is transparent and predictable
- GitHub can maintain reliable service without gating access
What’s Changing: Premium Requests to AI Credits
The Old System: Premium Request Units (PRUs)
Under the legacy model, each Copilot action consumed a number of “premium requests” based on a model multiplier. For example:
- GPT-4.1: 1x cost
- GPT-5.4: 6x cost
- Claude Opus 4.8: 27x cost
Your monthly allowance was fixed. Once exhausted, you either fell back to a cheaper model or lost access.
The New System: GitHub AI Credits
Starting June 1, 2026, usage is measured in GitHub AI Credits. Here’s what changed:
- What’s counted: Every token matters now—input tokens, output tokens, and cached tokens
- Cost model: 1 AI Credit = $0.01 USD
- Pricing transparency: Each model has a published per-token rate, so you can calculate costs exactly
- Included allowance: Every paid plan includes monthly credits matching the subscription price
- No fallback: When you exhaust included credits, you can’t fall back to a cheaper model. Instead, you either stop using Copilot or pay for additional credits
Plan Pricing: What Stays the Same
This is important: subscription prices are not changing.
| Plan | Monthly Price | Included AI Credits | Notes |
|---|---|---|---|
| Copilot Free | $0 | $0 (limited usage) | Code completions only |
| Copilot Pro | $10 | $10 | Includes 1 million tokens at average usage |
| Copilot Pro+ | $39 | $39 | Includes 4-5 million tokens at average usage |
| Copilot Business | $19/user | $19 | Pooled across organization |
| Copilot Enterprise | $39/user | $39 | Pooled across organization |
For businesses and enterprises transitioning in June, GitHub is offering promotional included usage for three months:
- Copilot Business: $30 in monthly credits (instead of $19)
- Copilot Enterprise: $70 in monthly credits (instead of $39)
Understanding AI Credits and Token Costs
The key to understanding costs under the new model is understanding tokens and model rates.
What Are Tokens?
Tokens are the atomic unit of how language models process text. A token is roughly equivalent to 4 characters or 1 word. When you use Copilot:
- Input tokens: Text sent to the model (your prompt, code context, etc.)
- Output tokens: Text the model generates (suggestions, responses, generated code)
- Cached tokens: Context the model reuses or stores (e.g., a file you reference repeatedly)
How Credits Are Calculated
For each interaction, GitHub calculates:
Total Cost (in credits) =
(Input tokens × Input rate) +
(Output tokens × Output rate) +
(Cached tokens × Cached rate)All rates are per 1 million tokens. So a model with an input rate of $2.00 costs $0.000002 per input token.
Model Pricing Breakdown
Here are the most relevant models and their costs per 1 million tokens:
OpenAI Models
| Model | Input | Output | Cached |
|---|---|---|---|
| GPT-5 mini | $0.25 | $0.025 | $2.00 |
| GPT-5.4 mini | $0.75 | $0.075 | $4.50 |
| GPT-5.4 nano | $0.20 | $0.02 | $1.25 |
| GPT-5.4 | $2.50 | $0.25 | $15.00 |
| GPT-5.2 | $1.75 | $0.175 | $14.00 |
| GPT-5.5 | $5.00 | $0.50 | $30.00 |
Anthropic Models (Claude)
| Model | Input | Output | Cached Write | Cached Input |
|---|---|---|---|---|
| Claude Haiku 4.5 | $1.00 | $0.10 | $1.25 | $5.00 |
| Claude Sonnet 4.6 | $3.00 | $0.30 | $3.75 | $15.00 |
| Claude Opus 4.8 | $5.00 | $0.50 | $6.25 | $25.00 |
Google Models (Gemini)
| Model | Input | Output | Cached |
|---|---|---|---|
| Gemini 3.5 Flash | $1.50 | $0.15 | $9.00 |
| Gemini 2.5 Pro | $1.25 | $0.125 | $10.00 |
| Gemini 3.1 Pro | $2.00 | $0.20 | $12.00 |
Important: Code completions and next edit suggestions remain free and unlimited across all paid plans. They do not consume AI credits.
What Individual Users Need to Know
Copilot Pro Users ($10/month)
- You get $10 in AI credits per month
- That equals approximately 1 million tokens at average usage
- Realistic estimate: 50-100 substantial Copilot Chat interactions per month
- Once credits are exhausted, you cannot use paid features until next billing cycle
- Migration: Monthly Pro users auto-migrate on June 1, 2026
Copilot Pro+ Users ($39/month)
- You get $39 in AI credits per month
- That equals approximately 4-5 million tokens at average usage
- Realistic estimate: 200-500 substantial Copilot Chat interactions per month
- Heavy agentic users may exhaust credits if using complex reasoning models repeatedly
- Migration: Monthly Pro+ users auto-migrate on June 1, 2026
- Annual users: Your annual plan continues under the old request-based model until expiration. Model multipliers increased on June 1
Free Tier Changes
- Code completions remain free
- Chat and other premium features become more restricted
- GitHub is gradually loosening limits as the new billing system stabilizes
What Teams and Businesses Need to Know
Copilot Business ($19/user/month)
- Each user gets $19 in monthly credits
- Credits are pooled at the organization level—unused credits don’t disappear, they can be shared
- June promotional period: $30 per user per month (50% bonus for 3 months)
- Admins get new budget controls to set limits at enterprise, cost center, or user levels
- When pooled credits are exhausted, organizations can choose whether to allow overage at published rates or cap spending
Copilot Enterprise ($39/user/month)
- Each user gets $39 in monthly credits
- Credits are pooled at the organization level
- June promotional period: $70 per user per month (79% bonus for 3 months)
- Advanced budget controls for large organizations
- Organizations can track usage by user, cost center, or repository
New Admin Features
GitHub is rolling out new tools for organizations:
- Budget controls: Set spending limits at multiple levels
- Pooled usage: Share credits across teams instead of siloing them per user
- Billing reports: Track usage and costs with granular visibility
- Preview bill: See projected costs before June 1 (preview available in May)
Practical Cost Examples
Let’s walk through realistic scenarios to estimate costs.
Example 1: Quick Copilot Chat Session
Scenario: Developer asks Copilot to explain a code snippet and generate a unit test.
- Prompt + context: 500 input tokens
- Response: 300 output tokens
- Model: GPT-5.4 (at $2.50 per 1M input, $0.25 per 1M output)
Cost calculation:
Input cost: (500 ÷ 1,000,000) × $2.50 = $0.00125
Output cost: (300 ÷ 1,000,000) × $0.25 = $0.000075
Total: $0.001325 ≈ 0.13 AI CreditsExample 2: Extended Agentic Session
Scenario: Copilot Agent iterates on a complex bug fix across 3 files over 15 minutes.
- Prompt + context: 50,000 tokens (full files, error logs, conversation history)
- Response: 5,000 tokens
- Model: Claude Opus 4.8 (at $5.00 per 1M input, $0.50 per 1M output)
Cost calculation:
Input cost: (50,000 ÷ 1,000,000) × $5.00 = $0.25
Output cost: (5,000 ÷ 1,000,000) × $0.50 = $0.0025
Total: $0.2525 ≈ 25.25 AI CreditsExample 3: Daily Usage Pattern (Team)
Scenario: Team of 5 developers, each doing 2-3 Copilot Chat sessions per day with code review and agentic work.
- Average session cost: 10-50 AI credits (varies by model and session length)
- Daily team cost: 100-750 AI credits
- Monthly team cost: 3,000-22,500 AI credits
- Team budget: 5 × $39 (Pro+) = $195 in credits
Result: Heavy users on the team might exhaust monthly credits mid-month and need to purchase additional credits or reduce usage.
How to Estimate Your Usage
GitHub is providing tools to estimate your costs before June 1:
- Preview bill experience: Available on your Billing page in May
- Historical data: GitHub shows your projected costs based on your recent usage
- Model selection: The models you use most influence cost (heavier models like Claude Opus or GPT-5.5 cost more)
- Session length: Longer agentic sessions consume more tokens and cost more
Tips to Control Costs
- Use lighter models for simple tasks: GPT-5 mini and Claude Haiku 4.5 are 10-20x cheaper than powerful models
- Enable auto model selection: You get a 10% discount automatically
- Cache context when possible: Cached tokens cost 4-5x less than regular tokens
- Monitor usage: Use GitHub’s billing reports to track consumption by user and team
- Set budget caps: Use admin controls to prevent unexpected bills
FAQ
Does subscription price change?
No. Copilot Pro remains $10/month, Pro+ remains $39/month, Business remains $19/user/month, and Enterprise remains $39/user/month.
What if I run out of credits?
You cannot fall back to a cheaper model. Instead, you either wait until next billing cycle, purchase additional credits, or stop using premium Copilot features. Code completions remain free.
Are code completions affected?
No. Code completions and next edit suggestions remain free and unlimited on all paid plans. They do not consume AI credits.
What happens to my annual plan?
If you’re on an annual Copilot Pro or Pro+ plan, you stay on the old request-based model until your plan expires. Model multipliers increased on June 1, 2026. When your plan expires, you’ll transition to Copilot Free with the option to upgrade to a monthly plan.
Can I purchase additional credits?
Yes, if you subscribe to a monthly paid plan. However, if you subscribe through GitHub Mobile (iOS or Android), additional credit purchases are not available.
How are cached tokens cheaper?
Cached tokens are context that models can reuse or reference without reprocessing. This reduces compute cost, so GitHub charges significantly less (typically 4-5x less than regular tokens).
What about Copilot code review?
Copilot code review consumes both AI credits (for the model processing) and GitHub Actions minutes (for the agentic infrastructure). Both are billed separately.
How much will my team’s bill increase?
It depends on usage patterns. Plan prices stay the same, so worst case is if your team vastly increases Copilot usage. But the June promotional credits give you room to test without extra cost.
How are you planning for the transition to usage-based billing? Are you concerned about controlling costs for your team, or do you think this pricing model is fairer? Share your thoughts in the comments—your perspective might help other teams navigate the change.