VS Code extension · v1.1.0 · free

Know which GitHub Copilot model to use, before you waste a prompt on it.

A model can be enabled for your account but slow right now — or its provider can be mid-outage. Model Advisor ranks your GitHub Copilot models by health, latency, and task fit, inside VS Code.

The Model Advisor sidebar in VS Code showing a ranked list of GitHub Copilot models with latency bars and health indicators.

What it does

A compact instrument, not another dashboard

Open the sidebar, run a check, know which model to use. That's the whole loop — recommendation first, details second.

zero tokens

Token-free health checks

The default check never sends a prompt to any model. It reads public status feeds from OpenAI, Anthropic, Google, Moonshot AI, and GitHub, then scores every GitHub Copilot model enabled for your account — including Kimi, Microsoft MAI, and xAI Grok models.

opt-in

Explicit latency benchmarks

When speed matters, benchmark one model or all of them. It measures time to first token, saves recent history, and shows cache age, trend, and median latency.

ranked recommendation

One score, honest penalties

Every model starts at 100 and loses points for slow first tokens, stale cached latency, degraded providers, and active incidents. Confidence labels make uncertainty visible.

The Model Advisor sidebar: a recommended model card on top, a token-use notice, task-fit picks, and ranked model rows with latency chips and scores.
task fit

Light · medium · complex picks

Separate suggestions for quick edits, everyday coding, and deep reasoning — built only from the models your account can actually use.

always visible

Status bar summary

A compact, clickable summary stays in the status bar while you work, with full diagnostics in the output channel when you want them.

How it works

Built on the official VS Code model API

Models are discovered with vscode.lm.selectChatModels({ vendor: "copilot" }), so the list always matches what your account, organization, and session actually allow.

Detect

Lists your enabled GitHub Copilot models, groups them by inferred provider, and collapses duplicate aliases.

Check health

Reads public status pages for each provider plus the GitHub Copilot service — no prompts, no tokens.

Benchmark (optional)

Measures time to first token and cancels the stream on the first chunk. Cached benchmark results are shared across your VS Code workspaces and marked stale after your configured TTL.

Recommend

Combines latency, provider health, and incidents into a ranked list, with the best current option on top.

No hidden cost. Health checks use zero GitHub Copilot tokens. Only the explicit benchmark sends a prompt (hi by default), and it always shows a confirmation with exactly which models will receive it. If cached latency exists, you can reuse it instead, or disable Benchmark All entirely.

Signals

Where the health data comes from

All sources are public status feeds — nothing private leaves your editor.

SignalSource
OpenAIstatus.openai.com
Anthropicstatus.claude.com
Googlestatus.cloud.google.com
Moonshot AI (Kimi)status.moonshot.cn
Microsoft (MAI)githubstatus.com — no dedicated public feed, tracked via its delivery route
xAI (Grok)githubstatus.com — no machine-readable feed, tracked via its delivery route
GitHub Copilot servicegithubstatus.com

Get started

Up and running in under a minute

Install GitHub Copilot Chat, sign in to GitHub, then add Model Advisor. Open the sidebar icon and click Health check for an instant, token-free signal.

Install for VS Code