Search topic
github copilot billing 2026 cost
Practical outcome
Copilot's June-2026 switch to credit billing — what it costs now and how to track it.
Quick answer
Under the old system, a chat or agent request cost a fixed number of “premium requests” based on the model's multiplier, however much work the request did. GitHub announced on April 27, 2026 that this would end. A preview bill appeared on the Billing Overview page in early May, and the switch took effect on June 1, 2026. Base plan prices stayed the same. What changed is that a long agent run with a large context now costs in proportion to its tokens, so heavy agent users and light chat users on the same plan can see very different bills. Checked against GitHub's docs and changelog on November 8, 2026.
What changed with GitHub Copilot billing in 2026?
Plan fee
unchanged price
Included credits
resets on the 1st
Additional usage
$0.01 / credit
Invoice
base + usage
Other changes landed at the same time or soon after. The Copilot Max plan ($100 per month, 20,000 credits) launched on June 1. Organizations got user-level budgets. Copilot code review started consuming GitHub Actions minutes on June 1, on top of its credits. Since September 16, 2026, Business and Enterprise users can request a budget increase from their admin. Existing Business and Enterprise customers received promotional extra credits ($30 and $70 per month) for June to August 2026. That promotion has ended, which is one reason some org bills rose in September.
What is an AI credit, exactly?
An AI credit is one US cent of model usage. GitHub publishes a per-token price for every model (input, cached input, cache write and output), and a request costs the sum of its tokens at those prices. A selection of the current rates per 1M tokens:
- 1
Pick a feature
Chat (any mode), Copilot CLI, the cloud agent, Spaces, Spark, third-party agents and code review all use credits. Inline completions and next edit suggestions do not. - 2
A model answers
The tokens are priced at that model's rate. An agent run that reads many files and loops through tools sends the growing context again on every step. Cached input is cheap, but fresh input and output add up. - 3
Credits tick down
Your monthly pool shrinks. When it hits zero, an individual needs an additional-spending budget to continue. In an organization, paid usage is on by default unless an admin turns off the “AI credits paid usage” policy.
Inline completions are not the culprit
The grey-text suggestions you accept while typing are not billed in AI credits on paid plans. Free has a 2,000-completion cap. If your bill jumped, look at agent runs, the cloud agent, code review and frontier-model chat.Why did my Copilot bill go up?
Three things stack 🔥. First, cost now follows tokens: a long agent session that used to be one premium request per prompt is now charged for every token of context it re-sends. Second, the fallback is gone. Running out used to mean slower answers on an included model, and now it means paid usage or a hard stop. Third, model choice multiplies everything: the same turn costs about 4× more on Claude Opus 5.5 than on GPT-6.1 Terra, and about 100× more on GPT-6 Astra than on GPT-6 Luna. For organizations, the end of the June–August promotional credits and code review's new Actions-minutes charge add to this.
- Agent-mode / CLI loops56%
- Frontier-model chat22%
- Code review12%
- Standard chat10%
How do I estimate my Copilot cost?
Estimate tokens per model, price them, subtract the allowance, and the remainder is your additional usage:
Here tokensm,c is your token count for model m in class c (input, cached input, cache write, output), rm,c is GitHub's rate, and Aincl is your plan's included credits in dollars ($15 on Pro, $70 on Pro+, $200 on Max). Worked example: 20 agent turns a day on Claude Sonnet 5.5, each sending 60K tokens (50K cached) and writing 4K, costs $0.065 per turn, so $1.30 a day. Over 22 workdays that is $28.60. That is 2,860 credits, so a Pro user would pay about $13.60 in additional usage, and a Pro+ user would be well inside the allowance.
How do I track Copilot usage and cost?
GitHub's AI usage page (github.com/settings/billing → AI usage) is the source of truth for billing, and the Copilot status-bar icon in your IDE shows what is left. Copilot CLI's /usage command shows credits and tokens for the current session. If you run Copilot CLI alongside other agents, use npx whoburnedmore for one view that totals token usage across all of them. It reads Copilot CLI's local session files under ~/.copilot/session-state/ and only sends daily aggregates off the machine. With --local, nothing leaves at all.
$ npx whoburnedmore --local Counting your token usage… Local dashboard: file:///Users/you/.config/whoburnedmore/dashboard.html Re-run `npx whoburnedmore --local` to refresh it. Nothing left your machine.
The dashboard shows tokens by tool and by model, with input, output and cache splits. Multiply Copilot's per-model tokens by GitHub's rates to estimate credits. The estimate will not exactly match GitHub's bill (cache writes, long-context tiers and promotions differ), and it only covers Copilot CLI on this machine. Chat in the IDE, the cloud agent and code review are not in local files.
Watch the model, not just the count
Because per-token rates differ by up to 100×, the fastest saving is usually switching the default model for routine work rather than doing less. Track usage first, then move the defaults that actually move the bill. See how to check Copilot usage for every place the meter appears.What GitHub's dashboard shows vs what whoburnedmore adds
GitHub's page is authoritative for credits across every Copilot surface, but it only covers Copilot. whoburnedmore is not a billing source. It puts Copilot CLI next to every other agent you run, so you can compare and prioritise across the whole stack.
AI credits replaced premium requests
per AI credit
to total Copilot CLI vs other agents
Sources checked November 8, 2026: GitHub blog, April 27, 2026, changelog, June 1, 2026, changelog, September 16, 2026, billing for individuals, billing for organizations, models and pricing.
Related guides
How to Check GitHub Copilot Usage and AI Credits
Premium requests are now AI credits; check your allowance and track Copilot CLI tokens.
AI Coding Cost: Claude Code vs Codex vs Gemini
A real cost comparison of the big three — measured from your logs, not marketing.
How to Check Your AI Coding Token Usage
The cross-tool overview: one command that totals your token usage and cost across every AI coding agent you run.