Copilot Fusion Reaches the Free Plan, Minus the Effort Levels

GitHub made Copilot Fusion generally available on 28 October 2026, and its documentation states the feature is available with all Copilot plans, Copilot Free included. The limit is not access but depth: Fusion Standard is the only effort level reachable from the free plan, and the 30 per cent discount on model costs that paid plans receive through 30 November 2026 is a paid-plan benefit too.¹

This is a routing system rather than a model, which makes it easy to misread as a free upgrade. Copilot picks which models to use for each prompt and how many of them to involve. The interesting question is not whether free users get it, because they do, but which parts of it actually work without a subscription.

What Fusion does, and why it is not a model

You select Copilot Fusion in the model picker and choose an effort level. You are not choosing a model. You are choosing a system that decides, per prompt, which model or models to use and in what sequence.

GitHub documents three execution patterns. Fusion picks one for each request, so two prompts in the same conversation can be handled differently.²

Prices measured on 27 October 2026. Vendor plan terms, allowances and prices change without notice, so treat every figure on this page as a dated snapshot rather than a permanent fact. Where a number here differs from the vendor's own pricing page, the vendor's page is correct and the figure above is simply older than it.

The effort level sets the quality bar Fusion aims for. Fusion Standard balances cost, quality and latency for everyday work. Fusion High allows more orchestration for multi-file changes. Fusion Max prioritises the most capable models for the hardest reasoning. A fourth level, Fusion Fast, is not selectable and only appears in VS Code inline chat.

The efficiency claim is specific. GitHub says 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, using roughly half the tokens and about 40 per cent less time per task.

That is a claim about tokens, not about price per token. Roughly identical per-token pricing with half the tokens consumed is how the saving appears. Whether it holds outside a benchmark is the question December will answer.

The free plan's actual slice

GitHub's plan comparison states the free plan's model access as "auto model selection only", which is what Fusion is. The free tier was already using a router; Fusion is a better one with a user-facing effort dial that, on the free plan, has a single setting.

The discount is the more concrete exclusion. Thirty per cent off model costs through 30 November is worth real money to a Pro or Pro+ user hammering Fusion High on multi-file work. On the free plan there is no discount to apply, and the allowance is described only as "an allowance of GitHub AI Credits" without a published number.

The real free-tier constraint is credits, not access

Copilot moved to credit-based billing, and Fusion does not change that. Every model Fusion calls is billed at that model's standard rate. Fusion High and Max can call several models per task, so a single request can consume more credits than asking one model directly.

GitHub is explicit that Fusion only adds model calls when a task warrants them, and that a straightforward task does not pay for a review it does not need. That is a sensible design and it is also an unquantifiable one for a free user, because there is no published figure for the free allowance to measure against.

The practical consequence is a habit rather than a setting. On a free plan, Fusion Standard is the only option, it draws on the same credit pool as everything else, and the way to make it last is to keep prompts scoped. A vague request that triggers orchestration spends credits a specific one would not.

Why shipping to Free on day one is the story

Almost nothing this month reached the free tier on launch day. ChatGPT's iCloud connectors did, and that was notable enough to write about. Suno's MCP server nominally did, except the assistant you connect it to has its own paywall. Google's Jules free tier is being shut down entirely.

GitHub's decision is the opposite of Google's. Where Google is consolidating agent products upward into Antigravity and closing the free door, GitHub shipped its new orchestration layer to the $0 plan and paywalled the knobs instead of the feature.

That is a defensible strategy. Fusion Standard on free is genuinely useful for everyday edits, it costs GitHub little at that effort level, and it builds the habit before anyone pays. The paywall sits exactly where the expensive behaviour begins, which is where it should sit.

What happens on 1 December

The discount ends on 30 November 2026. Fusion itself does not, and the pricing model does not change: still per model, at standard rates, no Fusion surcharge.

What changes is the cost of the two premium effort levels. Anyone who adopted Fusion High or Max during the launch window and built a workflow around it will see model costs rise by the discount amount on their next cycle. GitHub has not said the discount will be extended.

For a Pro subscriber on $10 a month with 1,500 credits, that is worth measuring now rather than in December. Run a representative week at the effort level you actually use, check the credit consumption, and decide whether High is a habit worth keeping at full price or a launch-window indulgence.

Frequently asked questions

Is Copilot Fusion available on the free plan?
Yes. GitHub's documentation says Copilot Fusion is available with all Copilot plans, and the effort availability table lists Fusion Standard for Copilot Free.

Can free users select Fusion High or Max?
No. Those effort levels are available on paid plans, with Pro+ and Max reaching the full set.

Does Fusion cost extra?
No separate charge. You are billed for each model it uses at that model's standard rate, out of your plan's AI credits.

What is the 30 per cent discount?
Paid Copilot plans receive 30 per cent off model costs when using Fusion through 30 November 2026. The free plan does not receive it.

Does Fusion replace auto model selection?
Yes. Existing auto, HydraFusion and HydraFusion Max selections transition to the corresponding Fusion effort levels.

Where can I use it?
VS Code, the GitHub Copilot app, Copilot CLI and JetBrains IDEs. Additional clients are listed as coming soon.

General availability date, effort levels, execution patterns, the benchmark claim and billing model verified in GitHub's 28 October 2026 changelog entry and the About GitHub Copilot Fusion documentation on 31 October 2026. Plan-by-plan effort availability and credit allowances from GitHub's Plans for GitHub Copilot documentation the same day.

Sources

1. GitHub changelog: Route and orchestrate models with GitHub Copilot Fusion: github.blog/changelog/2026-10-28-route-and-orchestrate-models-with-github-copilot-fusion. Used for: GA date, the three effort levels, execution patterns, the Coding Agent Index claim and billing.
2. GitHub Docs: About GitHub Copilot Fusion: docs.github.com/en/copilot/concepts/models/copilot-fusion. Used for: Effort-level availability by plan, the 30 November discount and Fusion limitations.
3. GitHub Docs: Plans for GitHub Copilot: docs.github.com/en/copilot/get-started/plans. Used for: Copilot Free model access and AI credit allowances by plan.

All sources are the vendor's own plan or documentation pages, checked on 27 October 2026. Plan terms change without notice, so confirm current limits before you rely on them.