Model selection is a key part of AI cost management and overall governance. Each AI variant has multiple models – for example Claude has Fable, Opus, Sonnet, and Haiku – and each is the better option for different tasks and outcomes. This is a balance of capabilities, speed, accuracy, and price and will vary across different requirements and different times.
Even when organisations & users are aware of model selection as a consideration, choosing which model to use and when can be difficult. Microsoft have introduced their answer to this with GitHub Copilot Fusion, where rather than the developer picking the best model to use – Copilot does. They say:
“When you select Copilot Fusion in the model picker, you aren’t selecting a specific model. You’re choosing a system that routes and orchestrates models for each task. There’s no need to switch models during a task or decide which model suits each step. For each prompt, Copilot Fusion determines which models to use and how they work together, while you choose the right level of effort.
Route and orchestrate models with GitHub Copilot Fusion – GitHub Changelog
How it works
Copilot Fusion comes with 3 effort levels:
Standard – Chooses versatile models for general use
High – Chooses models for more challenging tasks that require deeper reasoning
Max – Picks models for the most demanding tasks & complex reasoning
It then has 3 execution models that may orchestrate multiple models:
Controlling GitHub Copilot Fusion
Copilot Fusion only uses the models that are enabled in your enterprise and organization model policies. Furthermore enterprise administrators can:
- view what models Copilot Fusion picks from (Managing availability of models in your enterprise – GitHub Docs)
- independently set each model’s effective policy state (Enterprise managed settings – GitHub Docs)
- disable models for Copilot Fusion without disabling the model for everyone
How is GitHub Copilot Fusion billed?
There is no additional charge for Fusion. The main thing to be aware of is that, if it chooses to use multiple models, you will be billed for multiple models.
Through the month of November 2026, for users with a paid GitHub Copilot plan – there is a 30% discount on model costs when using Fusion.
I am interested to see if Fusion brings about a significant reduction in cost/improvement in performance for many organisations.
