HydraFusion reaches VS Code and the Copilot app
Use it for uncertain coding chats, not as a blanket replacement for Auto.
Short answerUse HydraFusion for uncertain Copilot tasks; keep Auto or known models for routine coding chat.
By JasonPublished Oct 1, 2026Last verified Oct 1, 20265 min read
GitHub Copilot users now have another model-picker option to evaluate: HydraFusion, a research preview that has moved from Copilot CLI into Visual Studio Code and the GitHub Copilot app. The practical question is not whether HydraFusion sounds more advanced than Auto, but whether any day-to-day coding work should move to it now. The supplied GitHub Changelog says HydraFusion is not a single model. It coordinates multiple models and can choose between three workflows: a single-model answer, a cascade that escalates only if a quality gate rejects the first draft, or a critique loop where a read-only reviewer from another model family checks the draft before one revision. That suggests a different fit from Auto, which GitHub describes as choosing a model per request rather than coordinating a workflow inside the turn. But the release is still a research preview, available only to Copilot Pro, Pro+, Business, and Enterprise users, with administrator approval needed for Business and Enterprise preview features. So the decision is mostly about task uncertainty, access, and tolerance for preview behavior—not about replacing your normal Copilot setup across the board.
What changed
According to GitHub Changelog, HydraFusion is now available in Visual Studio Code and the GitHub Copilot app after previously being available through Copilot CLI. In VS Code, GitHub says it appears in the Copilot Chat model picker on version 1.140 or later, including VS Code Insiders. If it is missing, users may need to enable `chat.copilot.hydraFusion.enabled`. In the GitHub Copilot app, GitHub says users should update to the latest version, enable HydraFusion in Settings, and then choose it from the model picker.
Access is not universal. GitHub says HydraFusion is available to Copilot Pro, Pro+, Business, and Enterprise users. For Business and Enterprise accounts, an administrator may need to allow preview features.
What HydraFusion is actually doing
GitHub’s key distinction is that HydraFusion is not presented as one model. It is a coordinator that selects a workflow and may use multiple models in one turn.
The Changelog lists three workflows:
- Single: one selected model handles the task directly.
- Cascade: an efficient model drafts first, then a quality gate decides whether to accept the output or escalate to a stronger model.
- Critique: one model drafts, a read-only critic from another model family reviews it, and the original drafting model revises once.
That makes HydraFusion different from Auto in GitHub’s framing. Auto chooses a model for the request. HydraFusion is meant to choose both the workflow and the model coordination pattern.
Where it may make sense
If you already use Copilot Chat in VS Code or the Copilot app, HydraFusion looks most relevant for tasks where you want more structure around the answer: debugging, reasoning-heavy code questions, code-generation prompts where a weak first pass would cost time, or tasks where an independent critique step might be useful.
GitHub also says this release adds clearer transparency, more frequent progress updates, and better status signals during longer tasks. That matters because multi-step AI workflows can feel stalled if the interface does not explain what is happening.
Where to be cautious
This is still a research preview. GitHub explicitly says HydraFusion is subject to change. That should keep it out of any “replace your workflow today” framing.
There is also an access and setup filter. You need an eligible Copilot plan. VS Code users need version 1.140 or later, or Insiders. Managed organization and enterprise users may depend on an administrator enabling preview features.
Decision frame
Do not switch everything to HydraFusion. Add it as an option.
Use HydraFusion when the task is ambiguous enough that workflow selection, escalation, or critique could be useful. Keep Auto or your usual model choice for routine edits, quick explanations, and low-risk prompts where speed and predictability matter more than orchestration.
Our read: HydraFusion is worth adding to your Copilot toolbox, but not worth treating as the new default for every coding prompt. The useful distinction is workflow selection. Auto picks a model for a request; HydraFusion can decide whether the request should be handled directly, escalated after a quality gate, or reviewed by a separate critic before one revision. That framing makes it more interesting for tasks where you are unsure whether a quick model answer is enough: debugging, multi-step code changes, or prompts where a second-pass review may catch a weak first draft. The caution is equally clear. GitHub labels it a research preview and says it can change. It may also require a setting toggle, a recent VS Code version, the latest Copilot app, or administrator permission in managed accounts. For indie builders and remote developers, the sane move is to enable it where available, route ambiguous or higher-stakes Copilot Chat work through it, and keep Auto or a familiar model for routine questions.
Use HydraFusion for uncertain Copilot tasks; keep Auto or known models for routine coding chat.
Based on GitHub’s announcement, HydraFusion is a useful new Copilot option for tasks that may benefit from model orchestration, escalation, or critique. It is not yet a blanket replacement for Auto because the feature is still labeled as a research preview and may require specific versions, settings, plans, or administrator approval.
Skip it for routine Copilot prompts, managed environments where preview features are disabled, or workflows that cannot tolerate behavior changes from a research preview.
Read next
Follow new articles
Email updates are not live yet, and we are not collecting addresses. To follow new articles, use the RSS feed.