1. What happened
In a changelog entry dated October 28, GitHub announced that Copilot Fusion is generally available. Instead of picking one model, you pick Copilot Fusion and Copilot decides, for each request, which models to use and how they work together. There are three execution patterns: Single, where one model solves the task directly; Cascade, where an efficient model drafts and a quality check decides whether to accept it or hand it to a stronger model; and Critique, where one model drafts, another reviews, and the original revises. You choose an effort level of Standard, High or Max (GitHub changelog).
On cost, GitHub writes that in the Artificial Analysis Coding Agent Index, Copilot Fusion Max scored 60, within two points of Claude Code with Opus 5.5 Max, at about half the average API cost per task, about half the tokens and about 40% less time. These are GitHub's own figures (GitHub changelog). Billing is at each model's standard rate for every model used, and users on paid plans get a limited-time 30% discount on model costs when using Fusion. It is available on all Copilot plans in Visual Studio Code, the GitHub Copilot app, Copilot CLI and JetBrains IDEs (same changelog).
2. Japanese coverage and enterprise controls
Cloud Watch reported on October 29 that it was announced in the first-day keynote of GitHub Universe 2026, building on the research preview Project HydraFusion from September, and that model fees when using Fusion are discounted by 30% through November (Cloud Watch). That article puts the cost reduction shown in the keynote at about 40% per task, which is worded differently from the "about half" in GitHub's own changelog. Because such figures depend on conditions, this article leans on GitHub's own wording. On November 2, Publickey summarized Copilot Fusion as a multi-model feature meant to balance performance and cost (Publickey).
For administrators, Copilot Fusion only uses models that are enabled in the enterprise and organization model policies. Enterprise administrators can see which models Fusion picks from, disable particular models for Fusion alone, and set Fusion as the default model (GitHub changelog). According to the official docs, Fusion has no context window of its own, and each step runs within the limits of the model that handles it (GitHub Docs).
3. Why it matters to small businesses
Our view Development outsourcing fees and in-house engineers' hours used to vary by how skilfully people chose and used AI. If a system now adjusts that automatically, we believe the amount of work a given budget buys will diverge further between agencies that use AI well and those that do not. A company that keeps hiring an agency whose quotes still assume the pre-AI way of working could be paying a premium for the same features six months from now. That said, GitHub's performance and cost figures are its own measurements and may not hold for your projects. Until you have tested it yourself, we would not sign a contract that depends on those numbers.
4. What small businesses should check now
- Ask your developers or web agency whether they use GitHub Copilot and, if so, on which plan
- Check in the contract whether AI usage (model fees) is included in the development fee or billed separately at cost
- Because Fusion bills per model used, ask the administrator whether monthly limits and the range of allowed models can be set
- The 30% discount is time-limited, so compare quotes on the cost after it ends
- Put in writing with your agency who reviews AI-written code, and how
Our outlook
Over the next six months to two years, we expect it to become standard in development tools that people set a budget and a quality level and let the system choose the best combination of models, rather than picking a model themselves. The same pattern will likely spread beyond coding to quotes, customer enquiries and document drafting, where you set a quality level and a cost cap and AI plans the steps. The gap will then open between companies that have already decided what to hand to AI, within what limit, and who checks the results, and those that have not, which will either find costs hard to predict as usage grows or never dare to use it. Now is the time to lay that groundwork.
Questions this article answers
What is GitHub Copilot Fusion?
It is a feature, generally available since October 28, 2026, that automatically chooses and combines AI models. Instead of picking one model you pick Fusion, and Copilot decides the best models and approach for each request. It is available on all Copilot plans, including in VS Code.
Will Copilot Fusion really lower development costs?
GitHub says that in a benchmark the average API cost per task was about half, but that is its own measurement. Billing is at each model's standard rate, and paid plans get a time-limited 30% discount. Whether it is cheaper for your projects needs to be tested.
Sources
- GitHub Changelog: Route and orchestrate models with GitHub Copilot Fusion (2026-10-28)github.blog
- GitHub Docs: About GitHub Copilot Fusiondocs.github.com
- クラウド Watch: 「GitHub Universe 2026」開幕、最適モデル自動選択機能「Copilot Fusion」のGAを発表cloud.watch.impress.co.jp
- Publickey: GitHub、「Copilot Fusion」正式リリース。GitHub Universe 2026publickey1.jp
Related Q&A
Related services
If you want to try this in your own business
To see how the topics in this article would apply to your own operations, and what it would cost, send us a message through the contact form.