Copilot Fusion vs picking models yourself: what it costs
Copilot Fusion lets you choose an effort level (Standard, High or Max) instead of a model. GitHub bills it per model used, in AI credits at $0.01 each, and takes 30% off Fusion's model costs on paid plans through 30 Nov 2026. Below: what each plan includes, what GitHub does and doesn't publish, and a calculator to compare Fusion with choosing models by hand.
Last verified: · plan prices, credit allowances, Fusion discount terms and per-token model prices re-read on docs.github.com and github.blog that day. GitHub can change any of them.
Plans and included credits
| Plan | Price | Included AI credits per month | Fusion effort levels you can pick (Fast is not selectable) | Source |
|---|---|---|---|---|
| Copilot Free | Free | An allowance of GitHub AI Credits. GitHub does not state the amount on the pages we checked. | Standard | Source: Docs: Plans for GitHub Copilot · Docs: Usage-based billing for individuals |
| Copilot Student | Free (verified students) | An allowance of GitHub AI Credits. The amount is not stated on the pages we checked. | Standard | Source: Docs: Plans for GitHub Copilot · Docs: Usage-based billing for individuals |
| Copilot Pro | $10 / month | 1,500 (1,000 base + 500 flex) ≈ $15 | Standard | Source: Docs: Plans for GitHub Copilot · Docs: Usage-based billing for individuals |
| Copilot Pro+ | $39 / month | 7,000 (3,900 base + 3,100 flex) ≈ $70 | Standard, High, Max | Source: Docs: Plans for GitHub Copilot · Docs: Usage-based billing for individuals |
| Copilot Max | $100 / month | 20,000 (10,000 base + 10,000 flex) ≈ $200 | Standard, High, Max | Source: Docs: Plans for GitHub Copilot · Docs: Usage-based billing for individuals |
| Copilot Business | $19 / granted seat / month | 1,900 per user a month, pooled across the organization (≈ $19) | Standard, High, Max | Source: Docs: Plans for GitHub Copilot · Docs: Usage-based billing for organizations and enterprises |
| Copilot Enterprise | $39 / granted seat / month | 3,900 per user a month, pooled across the organization (≈ $39) | Standard, High, Max | Source: Docs: Plans for GitHub Copilot · Docs: Usage-based billing for organizations and enterprises |
| Copilot access (eligible enterprises) | $0 with GitHub Enterprise Cloud; $10 / user / month as Copilot Standalone | No included allowance. Metered usage is billed from the first use. | Not listed by GitHub (its availability table has no Copilot access column) | Source: Docs: Plans for GitHub Copilot · Docs: Usage-based billing for organizations and enterprises |
What Copilot Fusion is
When you select Copilot Fusion in the model picker you aren't selecting a model. You're choosing a system that routes and orchestrates models at runtime. For each prompt it picks how the work is done: one model answers (Single), an efficient model drafts and a stronger one is called if a quality check says so (Cascade), or one model drafts while others critique it in parallel (Critique). You pick the effort level. GitHub changelog: Copilot Fusion GA (28 Oct 2026) · Docs: About GitHub Copilot Fusion
| Level | What it prioritizes | Typical use |
|---|---|---|
| Fusion Standard | Balances cost, quality and latency. | Everyday work. Copilot Fusion selects one model for each task. |
| Fusion High | Balanced orchestration with moderate token savings. | Substantial, well-scoped coding tasks, such as fixing a complex bug or making a change across several files, when additional model calls may be worth the extra time and AI credits. |
| Fusion Max | Frontier capability while still considering token efficiency. | Tasks requiring the deepest possible reasoning and most thorough analysis. |
It became generally available on 28 Oct 2026 in Visual Studio Code, the GitHub Copilot app, Copilot CLI and JetBrains IDEs. High and Max are listed for Pro+, Max, Business and Enterprise, not for Free, Student or Pro. Administrators choose which models Fusion may use. GitHub changelog: Copilot Fusion GA (28 Oct 2026) · Docs: About GitHub Copilot Fusion
How it is billed
- Premium requests are gone on monthly plans. Before 1 June 2026 each model interaction cost one premium request unit, scaled by a model multiplier. As of 1 June 2026 GitHub replaced that with usage-based billing in AI credits. Premium-request multipliers now apply only to Pro and Pro+ subscribers who stayed on an existing annual plan. This page therefore has no premium-request allowance or multiplier table, because they would not describe what a monthly subscriber pays. Docs: What changed with Copilot billing (legacy) · Docs: Model multipliers for annual plans (legacy) · GitHub changelog: Updates to Copilot billing and plans (1 Jun 2026)
- One AI credit is $0.01. Each interaction is priced in tokens (input, cached input and output) at the model's rate and converted to credits. Docs: Usage-based billing for individuals · Docs: Models and pricing for GitHub Copilot
- Extra usage costs the same $0.01 per credit. On individual plans you set a dollar budget for it, and GitHub may limit how much extra usage an individual can add. On Business and Enterprise, extra usage is on by default and an administrator can switch it off. Docs: Usage-based billing for individuals · Docs: Usage-based billing for organizations and enterprises · Docs: Models and pricing for GitHub Copilot · GitHub changelog: Updates to Copilot billing and plans (1 Jun 2026)
- Included credits do not carry over. They reset at 00:00:00 UTC on the first day of each calendar month. Docs: Usage-based billing for individuals · Docs: Usage-based billing for organizations and enterprises
- Business and Enterprise credits are pooled across the organization or enterprise. For example, 100 Business users share a pool of 190,000 credits. Docs: Usage-based billing for organizations and enterprises
- Code completions and next edit suggestions are not billed in AI credits and stay unlimited on paid plans. Docs: Usage-based billing for individuals · Docs: Usage-based billing for organizations and enterprises
- Fusion keeps the main conversation on the same model where it can, so cached tokens keep helping. Cached tokens are typically billed at 10% of the normal input price, and switching models in the middle of a session makes the next request rebuild the cache at full input price. Docs: Optimizing your AI usage · Docs: About GitHub Copilot Fusion
- If you use auto model selection in Copilot Chat or the cloud agent on a paid plan, GitHub gives a separate 10% discount on model costs. That is not Fusion, and the calculator does not apply it. Docs: Usage-based billing for individuals
- On Standard, Fusion may use a small local model you have installed (MAI Code 1.1 Flash or Qwen3.8 Flash Next, public preview) (experimental features must be enabled). GitHub's page does not say how this is billed. The calculator ignores this. Docs: About GitHub Copilot Fusion
- Copilot Max is available as an upgrade for existing Student, Pro and Pro+ subscribers. GitHub's 1 Jun 2026 post said new sign-ups for Student, Pro, Pro+ and Max were paused and would reopen in the coming weeks. We did not check whether that has changed, so see GitHub's plans page. GitHub changelog: Updates to Copilot billing and plans (1 Jun 2026) · Docs: Plans for GitHub Copilot
- Fusion is generally available in VS Code, the Copilot app, Copilot CLI and JetBrains IDEs. GitHub's launch post says support for additional clients is coming soon. GitHub changelog: Copilot Fusion GA (28 Oct 2026) · Docs: About GitHub Copilot Fusion
Fusion vs picking models: monthly cost calculator
Everything runs in your browser, nothing is sent anywhere, and all prices are embedded in this page. Model rates are GitHub's published per-token prices. The task sizes and the Fusion routing settings are our placeholders, so treat the answer as a what-if, not a quote.
What GitHub doesn't publish
These are the gaps the calculator has to cover with assumptions, so they matter most when you read the result.
- Which models Fusion picks at each effort level. GitHub says Fusion draws on a pool that changes as new models arrive and as it evaluates performance, and that you can hover over the effort name after a response to see the models used. It does not publish a mapping from effort level to models, or how often Cascade or Critique runs (we did not find either on the pages we checked). Docs: About GitHub Copilot Fusion
- What Fusion costs per task versus choosing a model yourself. GitHub publishes no cost comparison per effort level. It says High and Max may use more AI credits than Standard, and that Fusion bills each model it uses at that model's standard rate with no separate Fusion charge. Its only number is a GitHub-reported benchmark in the launch post: Fusion Max scored 60 on the Artificial Analysis Coding Agent Index, within two points of Claude Code with Opus 5.5 Max, at about half the average API cost per task. That compares Fusion with a different product, not with a model you pick inside Copilot. GitHub changelog: Copilot Fusion GA (28 Oct 2026) · Docs: About GitHub Copilot Fusion
- How the 30% discount interacts with your included credits. GitHub says paid plans get a 30% discount on model costs when using Fusion, through 30 Nov 2026. It does not say whether the discount shrinks the credits drawn from your allowance or only the price of extra usage. Our calculator assumes the first: credits used by Fusion are multiplied by 0.7. Untick the discount box to see the other case. Docs: About GitHub Copilot Fusion · Docs: Usage-based billing for individuals
- What happens after 30 Nov 2026. The docs say the discount runs through 30 Nov 2026. The launch post only says "for a limited time". We found no extension or later price on the pages we checked. Docs: About GitHub Copilot Fusion · GitHub changelog: Copilot Fusion GA (28 Oct 2026)
- Whether Copilot access seats get the discount. The docs say "paid Copilot plans". They do not list which plans that covers. Copilot access has a $0 or $10 seat charge and no included credits. Docs: About GitHub Copilot Fusion · Docs: Plans for GitHub Copilot
- Free and Student credit amounts. Both plans have "an allowance of GitHub AI Credits". The plans and billing pages we checked give no number, so the calculator leaves it for you to enter. Docs: Plans for GitHub Copilot · Docs: Usage-based billing for individuals
- The flex part of your allowance. Pro, Pro+ and Max include base credits that never change plus a flex allotment that GitHub says is variable and designed to adapt as the economics of AI evolve. The totals here are today's. Docs: Usage-based billing for individuals
- Cloud sandbox size and the exact start of sandbox billing. GitHub publishes the three sandbox rates but not how many vCPUs, how much memory or how much disk a cloud sandbox gets. It says billing for sandbox usage begins in November 2026 and gives no day on the pages we checked. The calculator's sandbox memory figure is therefore our assumption. Docs: Billing for cloud and local sandboxes · GitHub changelog: New connected Copilot cloud experience (28 Oct 2026)
Assumptions and method
- One "task" is one agent-style job with the token counts you enter. Defaults (150,000 fresh input, 600,000 cached input, 15,000 output tokens per task, 8 tasks a day, 21 working days) are illustrative, not GitHub figures. Use the AI usage page in your GitHub billing settings to get your own.
- Token cost per task = fresh input × input rate + cached input × cached rate + output × output rate, using the default-tier rates from GitHub's Models and pricing page. Cache-write charges and long-context tiers are not modelled.
- Picking models yourself: you choose an everyday model and a hard-task model and say what share of tasks go to the hard one.
- Fusion: the effort-level settings (share of tasks sent to the hard-task model, and extra review overhead from Cascade or Critique calls) are OUR PLACEHOLDERS, not GitHub figures. Change them to what you see in your own usage. "Neutral" copies your manual mix with no overhead, so only the discount is left.
- Overage is credits used above the included allowance, at $0.01 per credit. On individual plans, GitHub says you keep working past your allowance by setting a budget for extra usage. Monthly cost = plan price + overage. Our figures are before tax.
- Business and Enterprise are modelled for one seat. Real pools are shared across all users, so heavy and light users offset each other.
- The Fusion discount is applied to credits used, only on paid plans, and only for the months it runs. The second Fusion column shows the same month without it.
- Cloud sandbox line (optional, off at 0 hours): usage cost = hours × 3,600 × ($0.000024 + memory GiB × $0.000003). That charges one compute second per running second, as GitHub's worked example does. The part of that cost covered by your included compute credit is free, and the rest is billed. Snapshot storage ($0.005 per GiB-month) is added on top, because GitHub lists only sandbox compute and memory as drawing from the included amount. The same sandbox cost is added to both columns, because it is billed apart from AI credits. The memory figure is our assumption, and your included credit is shared with GitHub Actions runner use, so enter what you have left.
Sources
Every GitHub price, allowance, rate and date on this page comes from the official GitHub pages below, opened on .
- GitHub changelog: Copilot Fusion GA (28 Oct 2026), opened 29 Oct 2026
- Docs: About GitHub Copilot Fusion, opened 29 Oct 2026
- Docs: Optimizing your AI usage, opened 29 Oct 2026
- Docs: Plans for GitHub Copilot, opened 29 Oct 2026
- Docs: Usage-based billing for individuals, opened 29 Oct 2026
- Docs: Usage-based billing for organizations and enterprises, opened 29 Oct 2026
- Docs: Models and pricing for GitHub Copilot, opened 29 Oct 2026
- Docs: What changed with Copilot billing (legacy), opened 29 Oct 2026
- Docs: Model multipliers for annual plans (legacy), opened 29 Oct 2026
- GitHub changelog: Updates to Copilot billing and plans (1 Jun 2026), opened 29 Oct 2026
- Docs: Billing for cloud and local sandboxes, opened 29 Oct 2026
- Docs: Product usage included with each plan, opened 29 Oct 2026
- GitHub changelog: New connected Copilot cloud experience (28 Oct 2026), opened 29 Oct 2026
FAQ
What is GitHub Copilot Fusion?
Instead of picking a model, you pick Copilot Fusion in the model picker and then an effort level: Standard, High or Max. For each prompt it decides which models to use and whether one model answers directly (Single), an efficient model drafts and a stronger one is called if needed (Cascade), or one model drafts while others critique it (Critique). It became generally available on 28 Oct 2026 in VS Code, the Copilot app, Copilot CLI and JetBrains IDEs. GitHub changelog: Copilot Fusion GA (28 Oct 2026) · Docs: About GitHub Copilot Fusion
Does Copilot Fusion cost extra?
No separate charge. You pay each model it uses at that model's standard rate, in GitHub AI Credits. High and Max can use more than one model, so one task can cost more than on Standard. Docs: About GitHub Copilot Fusion
How much is the Fusion discount and when does it end?
30% off model costs for users on paid Copilot plans, through 30 Nov 2026 according to GitHub's docs. The launch post says "for a limited time" without a date. Docs: About GitHub Copilot Fusion · GitHub changelog: Copilot Fusion GA (28 Oct 2026)
Which plans can use High and Max?
Fusion Standard is included on every plan. GitHub's availability table lists High and Max for Pro+, Max, Business and Enterprise, and not for Free, Student or Pro. Docs: About GitHub Copilot Fusion
What does an overage cost?
$0.01 per AI credit above your included allowance, so a $10 budget covers 1,000 credits. Docs: Usage-based billing for individuals · Docs: Models and pricing for GitHub Copilot
Are premium requests and model multipliers still used?
Not for monthly plans. GitHub replaced premium requests with usage-based AI credits on 1 June 2026. Multipliers survive only for Pro and Pro+ subscribers who stayed on an existing annual plan. Docs: What changed with Copilot billing (legacy) · Docs: Model multipliers for annual plans (legacy)
Will Fusion be cheaper than choosing a model myself?
GitHub does not publish that comparison. Fusion Standard is meant to balance cost and quality, and until 30 Nov 2026 paid plans get a 30% discount on model costs when using Fusion. Whether Fusion is cheaper for you after that depends on which models it picks for your work. That is our reading, and only your own usage page can show it. Docs: About GitHub Copilot Fusion · GitHub changelog: Copilot Fusion GA (28 Oct 2026)
Can I stop Fusion from using a model?
Administrators control which models Fusion can use and can set Fusion as the default. Fusion never uses models excluded by an administrator's policy. GitHub changelog: Copilot Fusion GA (28 Oct 2026) · Docs: About GitHub Copilot Fusion
Do Copilot cloud sandboxes cost extra on top of AI credits?
Yes. GitHub bills cloud sandbox usage on separate meters: $0.000024 per compute second, $0.000003 per GiB-second of memory and $0.005 per GiB-month of snapshot storage. Compute and memory first draw on the compute included in your GitHub plan (for example $18 a month on GitHub Pro), shared with Actions runner use. GitHub says billing for sandbox usage begins in November 2026. The calculator's optional sandbox section adds this to both columns. Docs: Billing for cloud and local sandboxes · Docs: Product usage included with each plan · GitHub changelog: New connected Copilot cloud experience (28 Oct 2026)
Related
- GitHub Copilot: our review, tier and plans.
- Run coding agents on your own cloud bill: per-token billing through AWS, Google Cloud or Azure.
- AI subscriptions: what $20 a month gets you across tools.
- AI usage limits: what each plan includes and what happens at the limit.