# GitHub Makes Copilot Fusion Generally Available; Independent Test Puts It Two Points Behind Claude Code at Half the Cost per Task

Times of AI Desk · 2026-10-29 · Products

[https://timesof.ai/2026/10/github-copilot-fusion-ga-model-router-half-cost](https://timesof.ai/2026/10/github-copilot-fusion-ga-model-router-half-cost)

> GitHub made Copilot Fusion generally available on Oct. 28. Instead of a single model, users pick an effort level and Copilot routes or combines Anthropic and OpenAI models for each task. Artificial Analysis scored Fusion Max at 60 on its Coding Agent Index, against 62 for Claude Code with Opus 5.5 at max effort, at $5.59 per task versus $11.42. Billing is per underlying model, with a time-limited 30% discount.

GitHub is betting that most developers would rather choose how hard a model should work than which model does the work, and an independent benchmark suggests that is a reasonable bet on cost. **Copilot Fusion**, generally available since **October 28**, replaces the model picker's choice of a single model with a system that decides, prompt by prompt, which models to use and how to combine them.

## How it works

Users choose an effort level: Standard for everyday coding, High for harder tasks and Max for the most demanding ones, GitHub says. Copilot then handles each task in one of three ways: a single model solves it; an efficient model drafts and a quality check decides whether to escalate to a stronger one (Cascade); or one model drafts, another reviews and the first revises (Critique). Fusion combines GitHub's earlier Project HydraFusion orchestration with its auto model selection.

The model pool includes **Claude Opus 5.5**, **Claude Sonnet 5.5**, **GPT-6 Astra** and **GPT-6 Sol**, according to Artificial Analysis. Customers are billed for each model Fusion uses at that model's standard rate, and paid plans get a 30% discount on Fusion model costs for a limited time. Administrators decide which models Fusion may use and can make it the default.

## The independent check

**Artificial Analysis** benchmarked Fusion Max in GitHub Copilot CLI on its Coding Agent Index v2.0. It scored **60**, against **62** for Claude Code with Opus 5.5 at max effort, at **$5.59 per task** versus **$11.42**. Devin Fusion, which pairs Opus 5.5 with GPT-6 Luna, also scored 62, at $8.00 per task. Fusion led the index's Terminal-Bench 4.0 component with 71 but ranked third on FrontierCode at 51 and scored 57 on SWE-Atlas-QnA-AA. It used 7.2 million tokens per task, about half as many as Opus 5.5 in Claude Code, Artificial Analysis says.

GitHub's own announcement adds that Fusion took about 40% less time per task; Artificial Analysis's write-up does not report timing. VentureBeat noted that a 77% figure GitHub cited came from a separate internal test, not the Artificial Analysis index.

For GitHub, which Microsoft owns, the selling point is now the orchestration rather than any one frontier model, which becomes a line item on the bill. Microsoft is building agent infrastructure on the same multi-model footing, as with the [Copilot harness in Foundry](https://timesof.ai/2026/10/microsoft-copilot-harness-foundry-agent-service-preview).

## Limits

- The Artificial Analysis result covers Fusion Max only; Standard and High were not independently scored.
- Cost per task depends on each model's list price, so it will change when Anthropic or OpenAI reprice.
- The 40% time saving and the AT&T efficiency comments are GitHub's and AT&T's own claims.

## Sources

- [GitHub Changelog: Route and orchestrate models with GitHub Copilot Fusion (October 28, 2026)](https://github.blog/changelog/2026-10-28-route-and-orchestrate-models-with-github-copilot-fusion/)
- [GitHub Docs: About GitHub Copilot Fusion](https://docs.github.com/en/copilot/concepts/models/copilot-fusion)
- [Artificial Analysis: GitHub Copilot Fusion orchestrates Anthropic and OpenAI models to balance performance and cost (October 28, 2026)](https://artificialanalysis.ai/articles/github-copilot-fusion-orchestrates-anthropic-and-openai-models-to-balance-performance-and-cost)
- [VentureBeat: GitHub Copilot Fusion mixes models to achieve near-Opus 5.5 coding performance at half the per-task cost](https://venturebeat.com/orchestration/github-copilot-fusion-mixes-models-to-achieve-near-opus-5-5-coding-performance-at-half-the-per-task-cost)