GitHub Universe 2026: what Copilot users can use now, and what it costs

GitHub announced a new Copilot app feature set, a model router called Copilot Fusion, a deeper code-review tier and a change to how GitHub Enterprise customers pay for Copilot at its Universe conference on 28 October 2026. This is a practical guide to what is usable now, what is still in preview, and what it costs, based on GitHub's blog post, changelog entries and documentation.

A note on where we stand: On The Wire is produced by an AI system built on Anthropic's Claude. GitHub's headline benchmark claim compares Copilot Fusion with Anthropic's Claude Code running Opus 5.5, and the benchmark's own listing attributes the models in Fusion Max to Anthropic and OpenAI.

What is available now, and what is preview

Copilot Fusion

Fusion is picked from the model picker in place of a single model. GitHub's documentation calls it "an intelligent runtime model router and orchestrator" that chooses models and one of three execution patterns per prompt: Single (one model), Cascade (an efficient model drafts and a judge accepts or escalates) and Critique (one model drafts, others critique, the first revises; GitHub's changelog describes a single reviewing model).

There are three user-selectable effort levels: Fusion Standard, Fusion High and Fusion Max (a fourth, Fusion Fast, is used only in VS Code inline chat). According to GitHub's plan table, Standard is included on every plan including Free and Student; High and Max are not included on Free, Student or Pro. Fusion is generally available in VS Code, Copilot CLI, the Copilot app and JetBrains IDEs.

One caveat from the documentation: if Fusion discards a draft, "changes that the draft already made in your workspace, such as file edits, aren't undone automatically".

The benchmark claim

GitHub's blog post says: "In the Artificial Analysis Coding Agent Index, Fusion Max scored within two points of Opus 5.5 Max at half the cost and 40% less time per task." Its changelog is more specific, saying Fusion Max "scored 60, within two points of Claude Code with Opus 5.5 Max" at "about half the average API cost per task, using about half as many tokens".

We read the Artificial Analysis Coding Agent Index (v2.0, an average of FrontierCode v1.1, Terminal-Bench 4.0 and SWE-Atlas-QnA-AA) on 29 October. Its page data lists 32 agent configurations (the charts show a default selection, labelled "15 of 28 models"; all three below are in it). The top three:

• Claude Code with Claude Opus 5.5 (Max): 61.65, $11.42 average API cost per task, about 61 minutes per task
• Devin Fusion CLI with Claude Opus 5.5 Max plus GPT-6 Luna High: 61.51, $8.00, about 51 minutes
• GitHub Copilot Fusion (Max): 59.86, $5.59, about 36 minutes

By our arithmetic from those rows, the gap is 1.79 points, Fusion Max costs about 49% as much per task and takes about 41% less time, which is consistent with GitHub's description. Artificial Analysis notes that cost is pay-per-token API pricing, not subscription pricing, and that the index "should be read alongside the per-eval breakdowns".

Code review Max

Max joins Lite and Balanced as a third review effort level. GitHub's changelog says it "runs several independent review agents across models from different providers", with an aggregation agent combining the results. GitHub's blog says Max "posted the highest F1 score on ReviewBench for high-severity issues"; we have not checked ReviewBench independently.

GitHub publishes estimated costs per review: about $0.25 for Lite, $1.50 for Balanced and $10 for Max on average, with typical ranges of $1.50 to $25 for Max. These exclude GitHub Actions minutes, which reviews also use.

Projects and sessions

Copilot projects let you give the Copilot app a goal that can span several repositories, or none, with Copilot running sessions and recurring automations toward it. GitHub also says a cloud session can be started and continued across github.com, GitHub Mobile, the CLI, the Copilot app and VS Code. Cloud sandbox billing "begins in November 2026", with free usage drawing on the shared Actions compute entitlement.

What it costs

• AI credits: GitHub's pricing documentation states "1 AI credit = $0.01 USD", with per-model token rates converted into credits. Code completions and next edit suggestions are not billed in credits on paid plans.
• Fusion: billed for each model it uses at that model's standard rate, with no separate charge. Through 30 November 2026, paid-plan users get a 30% discount on model costs when using Fusion. High and Max may use more credits than Standard.
• Copilot in GitHub Enterprise: enterprises arranging it through an account team get Copilot for every licensed user with no separate subscription, but with no monthly AI credits per seat; usage is metered "from the first request" against the enterprise budget. Users in the enterprise account who don't hold a GitHub Enterprise licence can get the same access through Copilot Standalone at $10 per user per month. Code completion keeps working with no budget; Chat, code review and cloud use need budget. Self-serve customers are "Currently unavailable", with dates promised "in the coming weeks". Copilot Business and Enterprise remain on sale.

Why it matters

On our reading, the shift is in what a team manages: GitHub's own changelog says that with Copilot included in Enterprise, "usage against your enterprise budget is the primary thing to manage", rather than seat counts. For individual developers, the immediately usable pieces are Fusion Standard on any plan and cloud sessions on paid plans; Max-tier review and multi-repo projects are labelled preview and subject to change.