Copilot: Language model unavailable and other errors, what to do

The messages "Language model unavailable" and "We are having troubles" in GitHub Copilot say that chat could not get an answer from the model, but do not explain why. There are usually several possible causes: an outdated editor or extension, a sign-in problem, an exhausted limit, an organization policy, the network, or an outage on the service side. Below is a checking order from the fastest to the longest, and a fallback so you do not sit idle.

We rely on the GitHub and VS Code documentation (the Troubleshooting and FAQ sections, as of October 18, 2026). The exact error text and its meaning depend on the extension version, so where we cannot confirm it we say plainly that you should look in the logs.

Step 1. See whether it is an outage

GitHub's documentation begins with one recommendation: if Copilot stops working, check the GitHub status page (githubstatus.com) for active incidents with Copilot or model availability. If there is a problem there, your settings are not at fault, and retrying in a few minutes may help.

Step 2. Update VS Code and extensions

This is the most common cause according to the documentation. Copilot Chat is tightly integrated with the VS Code UI, so each new version of the Copilot Chat extension is compatible only with the latest editor release. An old version cannot talk to Copilot servers or does not see new models.

What to do:

1. Run Code: Check for Updates in the Command Palette.
2. Update the GitHub Copilot and GitHub Copilot Chat extensions in the Extensions view.
3. Reload the window: F1, then Developer: Reload Window.

Step 3. Check sign-in and plan

If you are signed in to GitHub but Copilot is unavailable, an authentication problem is possible. The documentation suggests this order:

1. Click the Accounts icon in the bottom-left corner of the window, hover your GitHub username and click Sign out.
2. Reload the window (Developer: Reload Window).
3. Sign in again.

Also check:

• that you are signed in with the GitHub account that holds the Copilot plan (there can be several in VS Code);
• that the subscription is active in your Copilot settings on GitHub;
• that, if you use a plan through an organization, access has been granted to you.

An error like "Could not connect to server. Extension activation failed" means, according to GitHub's documentation, either that there is no Copilot plan or that fetching a token from api.github.com failed; signing out and in again helps.

Step 4. Limits and credits

The next thing to check is exhausted limits. Open the Copilot status panel in the VS Code status bar: it shows what percentage of the monthly suggestion and AI credit allowance has been used. Two things need to be told apart:

• the plan's monthly allowance: when it runs out, paid models in chat stop answering until the reset or until you configure an additional budget;
• a temporary rate limit: a limit on request frequency that can hit even when the monthly allowance is not exhausted, especially for models with limited capacity. The documentation advises waiting a few minutes and retrying.

If both limits are fine and the error stays, move on. We do not list plan terms because they change: see the current ones on GitHub's site.

Step 5. Organization policies and model choice

If Copilot is provided by an organization, an administrator can restrict the set of models and switch off individual features. The VS Code FAQ says directly that not all models are available in the picker and that organizations can restrict access to certain models. Check that you did not pick a model that is not in your plan in chat, and try switching to another from the list.

If agent mode is unavailable altogether, look at the chat.agent.enabled setting: it may be managed by an organization policy.

Step 6. Network, proxy and diagnostics

If Copilot cannot connect, the documentation advises collecting diagnostics:

1. Run GitHub Copilot: Chat Diagnostics in the Command Palette.
2. Study the report: failed checks point to a blocked connection.
3. Compare against firewall and proxy rules (GitHub's documentation has a list of addresses to allow).

If the report is clean, enable detailed logs: Developer: Set Log Level for the GitHub Copilot and GitHub Copilot Chat extensions, reproduce the error and open Output, the GitHub Copilot Chat channel. Error lines near the time of failure usually name the cause more precisely than a pop-up. Before sending logs to support, remove secrets and paths: logs can contain code fragments.

Fallback: chat on your own key

While you work out the cause, you can keep working without the Copilot service. According to the VS Code documentation, models with your own key (BYOK) work in chat without signing in to GitHub and without a plan. The setup in outline:

1. Create a fetch_... key on the API keys↗ page.
2. In VS Code open Chat: Manage Language Models, choose Add Models, then Custom Endpoint.
3. Enter the address https://api.zukko.pro/v1/chat/completions, the key and a model identifier, for example claude-sonnet-5-5 or gpt-6.1-sol.

According to the VS Code documentation, Custom Endpoint also supports the Anthropic Messages API. For Claude models you can pick that type and the address https://api.zukko.pro/v1/messages: in the Anthropic format the gateway counts cache reads, while in the OpenAI format all Claude input is billed at the input price. Check the exact configuration fields against the VS Code documentation.

The limits of this route are the same as for BYOK in general: inline suggestions as you type, semantic search and embedding-based features remain with the Copilot service. Chat and agent mode (if the model supports tool calling) will work.

The second option is the Cline or Kilo Code extensions, which work with your own key from the start (Roo Code was shut down on May 15, 2026, according to its documentation). The address, key and format for them are collected in the VS Code extensions guide↗. If you prefer the terminal, take Claude Code: the instructions are in the connection documentation↗.

The fallback channel is cheap: for example, gpt-6.1-sol costs $0.27 per 1M input and $1.32 per 1M output tokens as of October 18, 2026. A chat of 20 requests of 4000 input and 600 output tokens costs 0.08 × 0.27 + 0.012 × 1.32 = 0.0216 + 0.01584 = about $0.04. Prices change; the current list is GET /v1/models.

Quick summary

FAQ

What does Language model unavailable mean?

Literally, the language model is unavailable. GitHub's documentation has no separate breakdown of this phrase, so look for the cause with the checklist: outage, outdated version, sign-in problem, limit, organization policy. The Copilot Chat logs will give the definitive answer.

Copilot stopped working after a VS Code update. What now?

Update the Copilot extensions too, reload the window (Developer: Reload Window), then sign out of GitHub and sign in again. If that does not help, collect diagnostics with GitHub Copilot: Chat Diagnostics.

Can I keep working while Copilot is unavailable?

You can move chat to your own key via Custom Endpoint: no GitHub sign-in and no plan needed. Inline suggestions will remain unavailable.

Will this work from Russia?

Connecting through the Zukko gateway does not depend on a Copilot plan. Access from Russia goes through the gateway, with payment by SBP, MasterCard or crypto.