GitHub made Copilot Fusion generally available on October 28, 2026, as part of its Universe announcements. In a changelog post, GitHub said: "GitHub Copilot Fusion is generally available." The same day it also shipped GitHub Agentic Workflows as generally available and said GitHub Enterprise can include Copilot access for licensed users.

VentureBeat covered the launch on October 28, 2026, adding Artificial Analysis benchmark dollars and an interview with GitHub chief product officer Mario Rodriguez. Those benchmark figures are credited below to Artificial Analysis as reported by VentureBeat, not as GitHub's own measurements.

#What can developers do with Copilot Fusion?

Instead of picking a model for each task, GitHub says users "choose Copilot Fusion and let Copilot intelligently route and orchestrate which models it uses for you." Rodriguez told VentureBeat: "It's not a model. It's an orchestration engine at runtime."

Users pick an effort level. Rodriguez told VentureBeat: "You just enable it, and then you have three reasoning efforts: standard, high and max." GitHub's changelog describes three execution patterns Fusion may use: Single, Cascade and Critique. GitHub Docs define Critique as one model drafting while "multiple models independently critique it in parallel," then the original model revising.

GitHub Docs say the model pool changes over time and that users can hover the effort level name to see which models ran. VentureBeat reported that Artificial Analysis identified Claude Opus 5.5, Claude Sonnet 5.5, GPT-6 Astra and GPT-6 Sol in Fusion's pool. That is Artificial Analysis via VentureBeat, not a GitHub-published GA pool list.

#Where is it available?

GitHub says: "Copilot Fusion is available on all Copilot plans in Visual Studio Code, the GitHub Copilot app, GitHub Copilot CLI, and JetBrains IDEs. Support for additional clients is coming soon." VentureBeat restated the same client list.

GitHub Docs' effort-level table includes Fusion Standard across plans. Fusion High and Fusion Max are listed as included on Pro+, Max, Business and Enterprise, and not included on Free, Student and Pro. Research preview users of auto, HydraFusion or HydraFusion Max move automatically to the matching Fusion effort levels, GitHub says.

Billing: "When you use Copilot Fusion, you're billed for each model it uses, at that model's standard rate. There's no separate charge for Copilot Fusion," GitHub Docs say. High and Max can consume more AI credits than Standard.

#What do the benchmarks say?

GitHub's changelog states: "In the Artificial Analysis Coding Agent Index, Copilot Fusion Max scored 60, within two points of Claude Code with Opus 5.5 Max. It delivered that result at about half the average API cost per task, using about half as many tokens and taking about 40% less time per task."

VentureBeat reported the Artificial Analysis dollars in prose:

  • Fusion Max averaged $5.59 per task, versus $11.42 for Claude Code with Opus 5.5 Max.
  • Fusion's overall score was 60, versus approximately 62 for that Claude Code configuration.
  • Devin Fusion scored 62 at an average $8.00 per task.
  • Artificial Analysis measured 7.2 million tokens per task for Fusion, about half the Opus 5.5 Claude Code token use.

Component scores, also via VentureBeat: Fusion led Terminal-Bench 4.0 at 71, scored 51 on FrontierCode and 57 on SWE-Atlas-QnA-AA. Artificial Analysis tested Fusion Max through GitHub Copilot CLI on Coding Agent Index v2.0.

Keep these claims separate:

  1. Benchmark task costs and scores above are Artificial Analysis figures as reported by VentureBeat (and summarized by GitHub). They are pay-per-token benchmark costs, not a published Copilot enterprise invoice estimate. VentureBeat says so explicitly.
  2. About 40% less time is GitHub's claim. VentureBeat notes Artificial Analysis did not report the timing data behind it.
  3. Up to 77% savings is a different story. VentureBeat says Microsoft CEO Satya Nadella has promoted potential savings of up to 77%, but that figure refers to a separate company-reported DeepSWE evaluation, not Artificial Analysis's average. VentureBeat warns it should not be conflated with Artificial Analysis's approximately 51% average cost advantage against Opus 5.5.
  4. The 30% Fusion promo (next section) is a temporary subscriber discount, not the benchmark savings.

#What is the 30% discount?

GitHub Docs say: "Through November 30, 2026, users on paid Copilot plans receive a 30% discount on model costs when using Copilot Fusion." The October 28 changelog only says "For a limited time." VentureBeat calls it a temporary 30% discount and does not give the November 30 end date. The end date here is sourced to GitHub Docs, re-checked on November 6, 2026. If this article publishes after November 30, 2026, that line must be rewritten.

VentureBeat states that promotional reduction "is separate from the task-cost savings measured in benchmarks."

#Enterprise Copilot access and plan prices

GitHub says: "Enterprises working with a GitHub or Microsoft account team can now get Copilot included with GitHub Enterprise for every licensed user." That path has the same features and models as Copilot Enterprise but does not include monthly AI credits per seat. Usage is metered from the first request against the enterprise budget. Code completion and next-edit suggestions keep working with no budget; Chat, code review and cloud Copilot need budget.

Enterprise owners must still assign Copilot seats to individual members, enterprise teams or organizations, including users with a GitHub Enterprise Cloud license, according to GitHub Docs.

Users without a GitHub Enterprise license can use Copilot Standalone at $10 USD per user per month. As of November 6, 2026, GitHub Docs list Copilot Pro at $10 USD per month, Copilot Business at $19 USD per granted seat per month and Copilot Enterprise at $39 USD per granted seat per month. One AI credit equals $0.01 USD. GitHub Docs say: "Copilot is not currently available for GitHub Enterprise Server." Self-serve Enterprise Copilot inclusion is rolling out in stages, GitHub says.

#What else shipped at Universe?

GitHub also said GitHub Agentic Workflows, a project from GitHub Next, is generally available on github.com for Free, Team and Enterprise, with the default Copilot CLI engine requiring a Copilot license. Other Universe items include Copilot Memory GA on paid plans, shared Automations rolling out on Pro and above, and Copilot Projects and computer-use previews.

For other Copilot and coding-agent coverage, see Anthropic's FlashProve prototype and OpenAI's bug bounty raise.

Our analysis: Fusion's pitch is orchestration and task-level cost, not a new frontier model. The Artificial Analysis dollars VentureBeat published make GitHub's "about half" claim concrete, while also showing Fusion Max is not the top Index score. Readers should keep four numbers in separate buckets: AA benchmark cost, GitHub's 40% time claim, the DeepSWE/Nadella 77% figure, and the time-limited 30% promo ending November 30, 2026.

#What is unclear?

  • Official GA model pool. GitHub does not publish a fixed list; Artificial Analysis via VentureBeat is secondary identification.
  • How the 30% shows on invoices. No GitHub statement found.
  • Country and region matrix. Beyond Enterprise Server exclusion, no list.
  • Enterprise invoice examples. Neither GitHub nor VentureBeat publishes a worked AI-credit bill for a typical team.