GitHub’s Copilot Fusion Nears Opus 5.5 at Half the Cost

GitHub said on 28 October 2026, at its Universe conference, that Copilot Fusion is generally available. Fusion is a setting that routes each coding task to one or more AI models instead of asking the developer to pick one. 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, GitHub says in its changelog.

GitHub, which Microsoft owns, also said GitHub Enterprise customers can now get Copilot included for every licensed user and pay for the AI they use. Its Octoverse report adds that by September more than 40% of pull requests on GitHub started with an agent’s first draft, up from 17% in March. Microsoft reports quarterly results after the US market close on the same day.

Short explainer · AI code brief

A vertical news Short narrated from the published headlines and opening lines of this piece and three related FourWeekMBA pieces on AI models and pricing.

Business Pill · Model Routing

An evergreen Business Engineer explainer on model routing. It explains the general idea only and does not discuss GitHub or Copilot Fusion.

What Fusion Does and What It Costs

When a developer selects Copilot Fusion in the model picker, they are not choosing a model but “a system that routes and orchestrates models for each task,” the changelog says. The developer picks an effort level and Fusion picks the models.

• Three effort levels. Fusion Standard prioritises versatile models for everyday coding, Fusion High prioritises capable models for harder tasks, and Fusion Max prioritises the most capable models, according to GitHub.
• Three workflows. A single model can solve a task; in a cascade, an efficient model drafts and a quality check decides whether to escalate; in a critique, one model drafts, another reviews, and the first revises.
• Benchmark claim. GitHub says Fusion Max reached its score using about half as many tokens and about 40% less time per task than Claude Code with Opus 5.5 Max.
• Billing. Users are billed for each model Fusion uses, at that model’s standard rate. For a limited time, paid Copilot plans get a 30% discount on model costs when using Fusion.
• Availability. All Copilot plans, in Visual Studio Code, the GitHub Copilot app, the Copilot CLI and JetBrains IDEs, GitHub says.

Administrators decide which models Fusion may use and can make it the default. AT&T’s Hari Pokhrel, quoted in the changelog, said Fusion Standard “has improved token efficiency across AT&T by automatically selecting the right model for the task, reducing unnecessary spend and preserving capacity for higher-value work.”

What Was Posted on X

GitHub posted that in the Artificial Analysis Coding Agent Index, Copilot Fusion Max scored within two points of Claude Code with Opus 5.5 Max at about half the API cost per task, and offered 30% off through Nov. 30.

Rohan Paul posted that GitHub recorded 7.38 billion commits in September 2026, more than five times the figure a year earlier, and that GitHub has added 4 million CPU cores, 160 petabytes of high-speed storage and 50 Tb/s of network capacity so far this year.

Rohan Paul posted that GitHub recorded 7.38 billion commits in September 2026, more than five times the figure a year earlier, and that GitHub has added 4 million CPU cores, 160 petabytes of high-speed storage and 50 Tb/s of network capacity so far this year.

Copilot Moves Into the Enterprise Plan

Enterprises working with a GitHub or Microsoft account team can now get Copilot included with GitHub Enterprise for every licensed user, a second changelog says. Usage against the enterprise budget, rather than a count of seats, becomes the main thing to manage.

• Metered from the first request. AI usage is billed against the enterprise budget, and there is no included allowance to track or top up, GitHub says.
• No cut-off. Code completion keeps working with no budget left, while Copilot Chat, code review and cloud agents need available budget; an exhausted budget never removes access, according to the changelog.
• Users without an Enterprise licence. They get the same access through Copilot Standalone at $10 per user per month.
• Rollout. The plan is available now through sales teams for GitHub Enterprise Cloud, and GitHub says it is opening it to self-serve customers in stages. Copilot Business and Copilot Enterprise remain on sale, it says.

What GitHub’s Own Data Shows

GitHub’s Octoverse report, published the same morning, gives the usage behind the launches.

• Agent-started work. More than 40% of pull requests started with an agent’s first draft in September 2026, up from 17% in March: 110 million in September alone, about 6x March’s volume.
• Volume. Developers pushed code 17.9 billion times in the past 12 months, more than in 2024 and 2025 combined; commits reached 34.3 billion and merged pull requests 1.3 billion.
• Agent features. 55% of Copilot’s monthly active users now use agentic features, up from 35% a year ago; the Copilot cloud agent grew 230% year over year.
• Model spread. Paid Copilot users used 29 distinct models in May and 47 in September, and no single model carried more than 25% of requests that month; 63% of paying Copilot developers used auto model selection in September.
• Developers. GitHub says it now serves more than 225 million developers.

Behind the launches, GitHub has added 4 million CPU cores, 160 PB of high-speed storage and 50 Tb/s of network capacity so far this year, Jay Parikh, Microsoft’s executive vice president of CoreAI, wrote. He said GitHub expects to finish migrating to a multi-region deployment in Azure early next year.

Connecting the Coverage

Ahead of Universe, Microsoft added server-to-server authentication for GitHub Apps in Copilot, alongside its own MAI image and cyber models, as we reported on 28 October. Microsoft also offered voluntary buyouts to about 60,000 employees, as we reported on 26 October, and its researchers ran 1,024 coding agents without a manager, as we reported on 25 October. On routing, SAP’s finance chief described picking the cheapest capable model from more than 70, as we reported on 28 October, and Mistral began serving DeepSeek beside its own models, as we reported on 28 October; our same-day look at falling AI prices is here. What’s new: GitHub has made routing the product, and puts a cost figure on it.

The Structural Read

As we read it, the cost claim is the headline: GitHub says Fusion Max landed within two points of Claude Code with Opus 5.5 Max in the Artificial Analysis Coding Agent Index at about half the average API cost per task.

GitHub’s Octoverse data shows the routing already happening by hand: paid Copilot users used 47 distinct models in September and no single model carried more than 25% of requests.

The Enterprise change moves Copilot from a per-seat add-on to usage metered from the first request against an enterprise budget, according to GitHub’s changelog.

Three Implications

FOR DEVELOPERS Fusion is available on all Copilot plans in VS Code, the GitHub Copilot app, the Copilot CLI and JetBrains IDEs, GitHub says, with a limited-time 30% discount on model costs for paid plans.

FOR ENTERPRISE BUYERS With Copilot included in GitHub Enterprise, usage against the budget becomes the main thing to manage, and an exhausted budget never removes access, according to GitHub.

FOR MODEL MAKERS Octoverse shows no single model above 25% of developer-selected requests in September, and Fusion now decides which models each task uses, GitHub says.

The Business Engineer Lens

This story maps onto the Business Engineer framework The Agentic Harness War.

The framework argues that the human’s role relocates as agents take over execution: “As execution delegates, human effort moves to delegation, supervision, review, and integration, and the returns to domain expertise rise.”

As we read it, Fusion moves one more decision into the harness: according to GitHub, the developer sets the effort level and Copilot picks the models. The framework was published in July 2026 and does not discuss GitHub or Copilot Fusion.

What Is Not Established

The benchmark comparison is GitHub’s own description of a result in the Artificial Analysis Coding Agent Index; we could not read the index page itself. “Half the cost” refers to average API cost per task in that test, not to Copilot subscription prices. GitHub did not give a price for Fusion beyond the per-model rates and the temporary 30% discount, and it did not say how many customers have moved to the Enterprise-included plan. We did not contact GitHub or Microsoft.

The Bottom Line

GitHub made Copilot Fusion generally available on 28 October 2026 and says its Max setting scored within two points of Claude Code with Opus 5.5 Max on the Artificial Analysis Coding Agent Index at about half the average API cost per task. It also folded Copilot into GitHub Enterprise with usage-based billing, while its Octoverse data shows agents starting more than 40% of pull requests in September.