Jan Desktop + QCode on Windows 11: setup guide (11 screenshots)

Set up Jan Desktop 0.8.8 with QCode on Windows 11 in 11 screenshots: install Jan, add a custom provider, fetch gpt-6.2-sol, and send your first chat.

Updated 2026-11-23

Need an API key? Sign up →Endpoints & base URLs →

On This Page

Short answer: In Jan’s Settings → MODEL PROVIDERS, click the + button to open Add Custom Provider, pick OpenAI-compatible, set Provider name to QCode, set Base URL to https://api.qcode.cc/openai/v1, paste your cr_ API key, click Create, then refresh Models and select gpt-6.2-sol.

Item Fill in
API format OpenAI-compatible
Provider name QCode
Base URL https://api.qcode.cc/openai/v1
API key cr_xxxxxxxx
Recommended model gpt-6.2-sol

Checked on 2026-11-22 · Jan 0.8.8 · Windows 11

Follow the 11 screenshots below on Windows 11 to install Jan Desktop, skip the multi-GB local model download, connect your QCode API key as a custom provider, and receive your first reply from gpt-6.2-sol. You can also add a second provider for claude-opus-5-5 using the same QCode key.

  • Applies to: Windows 10 / 11 (64-bit) (Windows 11 UI)
  • Software version: Jan 0.8.8
  • Screenshots updated: 2026-11
  • Time needed: about 10 minutes
  • You will need: a QCode API key starting with cr_

Before the first screenshot, check that you have these ready:

  • A 64-bit Windows 10 or Windows 11 PC with internet access
  • At least 1 GB of available disk space for the 545 MB installer and the Jan app (you do not need to download a local model to use QCode)
  • Access to GitHub Releases (github.com) or jan.ai to download the installer, plus api.qcode.cc for chat requests
  • A QCode API key copied from your dashboard (cr_xxxxxxxx). No account yet? Sign up first, pick a plan on the pricing page, then create the key in your Dashboard.

Download and install Jan

Jan’s Windows download button links directly to GitHub Releases. Pick the tab that matches where you are:

Open jan.ai/download and click Windows → x64 for standard Intel and AMD PCs (choose ARM64 only on Snapdragon / ARM PCs). If the site is temporarily unreachable, download Jan_0.8.8_x64-setup.exe directly from GitHub Releases v0.8.8.

Alternative install options: install Jan from the Microsoft Store, or run winget install Jan.Jan in PowerShell.

There is no official China mirror for Jan. Open GitHub Releases v0.8.8 directly in your browser and download Jan_0.8.8_x64-setup.exe (about 545 MB) for standard Intel and AMD PCs. If the official page jan.ai/download times out or the download stalls, wait a few minutes and retry, or switch networks.

You can also search for Jan in the Microsoft Store or install it from PowerShell with winget install Jan.Jan (store and winget versions may lag behind GitHub Releases).

Copy your QCode API key

Open the QCode console in your browser and click ① Sign in to log in to your account. After signing in, open API Key Management, click the eye icon to reveal your cr_ key, and click Copy. Keep this key ready for the provider form later.

Sign in to the QCode console first, then reveal and copy your cr_ key from API Key Management

Done right you will see: after clicking ① Sign in and opening your dashboard, you have a single key starting with cr_ copied to your clipboard with no extra spaces.

Download the Windows installer

Open GitHub Releases v0.8.8 (or jan.ai/download) and scroll down to the Assets list. Click ① Jan_0.8.8_x64-setup.exe (545 MB) for standard Intel and AMD Windows PCs. Only click ② Jan_0.8.8_arm64-setup.exe (320 MB) if your PC uses an ARM64 processor.

GitHub Releases Assets list showing Jan_0.8.8_x64-setup.exe (545 MB) for standard Windows PCs and Jan_0.8.8_arm64-setup.exe for ARM64 PCs

Stuck? The Assets list is empty or the installer will not run

GitHub loads the Assets list after the rest of the page. Scroll to the bottom and wait a few seconds before reloading. If you double-click the installer and nothing happens or Windows says the app cannot run on your PC, you likely downloaded arm64-setup.exe by mistake — go back and download Jan_0.8.8_x64-setup.exe instead.

Done right you will see: ① Jan_0.8.8_x64-setup.exe finishes downloading to your Downloads folder at roughly 545 MB (or ② Jan_0.8.8_arm64-setup.exe on an ARM PC).

Run the Jan installer

Double-click Jan_0.8.8_x64-setup.exe in your Downloads folder. Jan installs for your current user into %LOCALAPPDATA%\Programs\Jan with no folder picker and no Next / Finish screens — a single Jan Setup window shows Installing, then closes and launches Jan automatically.

Jan Setup window displaying the Installing progress bar with no wizard prompts

Stuck? Windows asks for administrator permission

Jan itself does not require administrator rights. However, on a fresh Windows PC that does not yet have Microsoft Visual C++ 2015–2022 Redistributable (x64) installed, the installer downloads that runtime first and triggers one User Account Control (UAC) prompt. Click Yes to finish installing the runtime, and Jan Setup will continue automatically.

Done right you will see: the Installing bar completes, the Jan Setup window closes on its own, and the Jan desktop app opens.

Skip the local model download and open Settings

On first launch, Jan runs a 4-step welcome flow: click Let's start, wait for the local engine check and click Continue, then choose Deny (or your preference) on the telemetry screen. On STEP 4 OF 4 — Your first model, do not click ① Download if you only want to use QCode cloud models — click ② Settings in the left sidebar instead to skip straight to settings.

Jan first-run Step 4 of 4 (Your first model) highlighting the Download button to skip and Settings in the left sidebar

Stuck? I clicked Download by accident

You do not need to wait for the local model download to finish. Click Settings in the left sidebar right away to configure QCode; you can cancel or delete the local model later under Hub or Llama.cpp to reclaim disk space.

Done right you will see: skipping ① Download and clicking ② Settings opens the Settings view without downloading a multi-GB local model.

Open the custom provider form

In the Settings sidebar, locate ① MODEL PROVIDERS (which groups built-in LOCAL and REMOTE providers). Click the ② + icon button directly to the right of the MODEL PROVIDERS heading to add a custom provider.

Jan Settings sidebar highlighting MODEL PROVIDERS and the plus icon button on its right

Done right you will see: clicking ② + next to ① MODEL PROVIDERS opens the Add Custom Provider dialog.

Choose the API format and name the provider

In the Add Custom Provider dialog, select ① OpenAI-compatible under API format, and type QCode into ② Provider name (Provider name (e.g. my-llm-server)). Keep the dialog open for the next step.

Add Custom Provider dialog with API format set to OpenAI-compatible and Provider name set to QCode

Done right you will see: ① OpenAI-compatible is selected and ② Provider name reads QCode.

Enter the Base URL, paste your API key, and click Create

In ① Base URL (https://your-endpoint/v1), enter https://api.qcode.cc/openai/v1 — keep the /v1 suffix with no trailing slash. In the API key field (Paste your API key) right below it, paste your cr_ key from the QCode Dashboard, then click ② Create at the bottom right.

Add Custom Provider dialog with Base URL set to https://api.qcode.cc/openai/v1, API key masked, and Create button highlighted

Need an API Key?

If you do not have a QCode account yet, Sign up first. To get started without a monthly plan, buy a pack on the Pay-as-you-go packs page, then create your key in the Dashboard and copy the string starting with cr_. You can pay by credit or debit card (Visa / Mastercard) — see Payment methods.

Stuck? Requests fail with 404 or 401 after saving

Check ① Base URL: for OpenAI-compatible, it must be exactly https://api.qcode.cc/openai/v1 (omitting /v1 causes 404). If you get 401, clear API key and re-copy your cr_ key from the Dashboard without leading spaces or quotation marks.

Done right you will see: ① Base URL reads https://api.qcode.cc/openai/v1, the API key box shows your pasted key, and clicking ② Create saves the provider and opens the QCode provider page.

Fetch the QCode model list

After clicking Create, QCode appears under REMOTE in the left sidebar and opens automatically. At first, ① Models displays No model found (and a grey banner at the top notes This provider may not be fully supported... — you can ignore that banner). Click the refresh icon on the right side of the Models header (the middle icon between search and +) to load models from QCode, including ② gpt-6.2-sol.

QCode provider page under REMOTE highlighting the Models section and gpt-6.2-sol after refreshing the model list

Stuck? Models still says No model found after refreshing

Confirm that Base URL ends with /v1 (https://api.qcode.cc/openai/v1) and that your cr_ key has no extra spaces, then click the refresh icon again. You can also click the + button next to the refresh icon and enter gpt-6.2-sol manually.

Done right you will see: under ① Models, the list populates with QCode models and ② gpt-6.2-sol is visible.

Start a new chat and pick gpt-6.2-sol

Click New Chat (Ctrl N) in the left sidebar. Click Select a model at the top left of the chat window to open the model dropdown, find the ① QCode group, and click ② gpt-6.2-sol.

Top-left model picker dropdown in a new chat highlighting the QCode provider group and gpt-6.2-sol

Done right you will see: the dropdown closes after you pick ② gpt-6.2-sol under ① QCode, and the top-left model selector displays gpt-6.2-sol.

Type a test message and send it

Verify that the top-left picker shows gpt-6.2-sol. Click inside ① Ask me anything..., type a short message such as Hello, please introduce yourself in one short sentence, and click the ② orange arrow button at the bottom right (or press Enter).

Chat input box Ask me anything with a test greeting typed and the orange send arrow button highlighted at the bottom right

Done right you will see: your prompt is typed in ① Ask me anything... and clicking ② sends it immediately.

Confirm the first reply

Wait a few seconds for the model to respond. Your prompt appears in ① the user message bubble, gpt-6.2-sol streams its reply underneath, and ② the tokens/sec status line appears below the reply.

Completed chat exchange showing the user message bubble, a streamed reply from gpt-6.2-sol, and the tokens/sec readout underneath

Done right you will see: ① your message bubble, a normal answer from gpt-6.2-sol, and ② a tokens/sec metric underneath with no red error banner.

Verify it works

Your Jan setup is ready when all three checks pass:

  • Settings → MODEL PROVIDERS → REMOTE lists QCode, and its Models section includes gpt-6.2-sol.
  • A new chat with gpt-6.2-sol returns a normal reply within a few seconds and shows a tokens/sec metric underneath.
  • Your Dashboard (or probe.qcode.cc) records the request under your cr_ key.

All done

You now have Jan Desktop connected to QCode on Windows 11 and chatting with gpt-6.2-sol. From here you can:

Common errors

Error message Why What to do
No model found Right after clicking Create, Jan opens the new custom provider page without populating its model list Click the refresh icon on the right side of the Models header to fetch models from QCode, or click + to add gpt-6.2-sol manually
This provider may not be fully supported... Jan displays this grey notice on every custom provider that is not a built-in preset Ignore the notice; as long as Base URL and your cr_ key are valid, QCode works normally once you refresh Models
model_not_available_on_endpoint A Claude model (such as claude-opus-5-5) was requested on an OpenAI-compatible provider, or a GPT model was sent to an Anthropic-compatible provider Use OpenAI-compatible (https://api.qcode.cc/openai/v1) for gpt-6.2-sol, and add a separate Anthropic-compatible provider (https://api.qcode.cc/api) for claude-sonnet-5-5
404 The OpenAI-compatible Base URL is missing /v1 (https://api.qcode.cc/openai), or an Anthropic-compatible Base URL includes an extra /v1 (https://api.qcode.cc/api/v1) Set Base URL to https://api.qcode.cc/openai/v1 for OpenAI-compatible, or https://api.qcode.cc/api (no /v1) for Anthropic-compatible
401 The API key is missing, expired, or pasted with extra spaces or quotation marks Return to your QCode Dashboard, click Copy on your cr_ key, clear API key in Jan, and paste it cleanly

FAQ

How do I use Claude models in Jan?

Claude models use QCode’s Anthropic-compatible endpoint rather than the OpenAI-compatible endpoint. In Settings → MODEL PROVIDERS, click + again to add a second custom provider with API format set to Anthropic-compatible and Base URL set to https://api.qcode.cc/api (do not add /v1), paste the same cr_ key, refresh Models, and pick claude-opus-5-5.

Can I use QCode in Jan without downloading a local model?

Yes. On first launch, skip the Download button on STEP 4 OF 4 — Your first model and click Settings in the left sidebar. Jan works purely with QCode cloud models without storing a multi-GB local model on your drive.

What happens when my QCode API key expires?

Renewal behavior depends on the plan type:

  • Personal plans: an expired personal plan key cannot be renewed. Purchase a new plan, generate a new cr_ key in your Dashboard, and replace the API key value inside your QCode provider settings in Jan.
  • Pay-as-you-go packs: when a pay-as-you-go pack expires, buying another pack automatically restores the same cr_ key. You do not need to change anything in Jan.

How do I switch Jan’s interface language?

Open Settings → General → Language and choose your preferred language from the dropdown menu.

Can I switch to a closer QCode access point?

Yes. Replace api.qcode.cc in Base URL with asia.qcode.cc or us.qcode.cc while keeping the exact same path suffix (/openai/v1 for OpenAI-compatible, /api for Anthropic-compatible).

Optional: add Claude models (Anthropic-compatible)

If you want both GPT and Claude models inside Jan, keep your QCode OpenAI-compatible provider for gpt-6.2-sol and add a second custom provider for Claude:

Item Fill in
API format Anthropic-compatible
Provider name QCode Claude
Base URL https://api.qcode.cc/api
API key cr_xxxxxxxx (the same QCode key)
Recommended model claude-opus-5-5

Jan appends /v1/messages automatically for Anthropic-compatible providers, so Base URL must stop at /api (adding /v1 causes 404). After clicking Create, click the refresh icon next to Models to list claude-opus-5-5.

Alternative access points

All three QCode domains share identical model capabilities and differ only in network routing:

Region OpenAI-compatible Base URL Anthropic-compatible Base URL
Default (global) https://api.qcode.cc/openai/v1 https://api.qcode.cc/api
Asia https://asia.qcode.cc/openai/v1 https://asia.qcode.cc/api
North America https://us.qcode.cc/openai/v1 https://us.qcode.cc/api

Get started with QCode

One API key for Claude Code, Codex and more — billed per token, per-model prices on qcode.cc/models

Enterprise →