Native Windows

The Windows node runs directly on Windows 10 22H2 or Windows 11, x64. It uses one or more NVIDIA GPUs with compute capability 6.1 or better and requires the Windows driver. WSL, Docker and virtualization are not needed. The models it offers are the ones your cards can run, as listed in Hardware compatibility.

Install or update

Open 64-bit Windows PowerShell. Leave at least 12 GB free on the local data drive for the first model, or 54 GiB free for the largest campaign candidate. The installer may ask Windows permission to install Microsoft’s signed Visual C++ runtime.

Invoke-WebRequest -UseBasicParsing 'https://docs.katara.com/install-native.ps1' -OutFile "$env:TEMP\katara-native.ps1"
powershell.exe -NoProfile -ExecutionPolicy Bypass -File "$env:TEMP\katara-native.ps1" -DocsBase 'https://docs.katara.com'

The installer takes the newest version published for Windows. To install a particular one, add -Version with its number, for example -Version '0.3.020'. Rerunning the installer preserves your data folder, model cache, provider settings and delegated key. Stop any running node before updating, then reopen Katara Node from the Start menu. To use another local fixed drive, add -DataDir 'D:\KataraData'. A temporary network share is unsuitable for the active cache.

Test this GPU

Choose 1. Test this GPU to check a card before registering it. The test runs Qwen3 4B with synthetic prompts, on the CUDA build that suits the card, and the first run downloads about 3 GB. A card with less memory than the model needs, or an older compute capability, is told so before anything downloads. The results are saved as a ZIP under your data folder. Testing is optional; you can choose 3. Start serving directly.

Start serving

Open Katara Node and choose 3. Start serving. If setup is incomplete, the app walks you through it:

1. Create or open the delegated node key. Register its public service-key address in the Provider portal with your wallet. Enter your public provider wallet address when prompted. Existing registration is remembered and checked automatically.
2. Choose a model. Setup lists the models the selected GPUs can serve, with their download sizes.
3. Enter your input and output prices in USDC per million tokens. Press Enter to accept 0.10 and 0.20, or enter your own amounts.

The app saves the configuration, downloads any missing model files and starts serving. You do not need to create or edit a configuration file. Existing cached files are reused. Your wallet private key stays in your wallet.

Each model is served by llama.cpp at a fixed quantization and context, one request at a time. One model runs at a time; choose 2. Set up wallet, model, GPUs, prices and concurrency to change it after stopping the node. An installation set up before the current list keeps its earlier Qwen3 profile, which targets Blackwell GPUs with at least 90 GiB of memory.

Settings are saved under your data folder in setup.env; public registration details are in provider.json. The default data folder is %LOCALAPPDATA%\Katara\data. The app prints the folder path. Enter Q during setup to cancel without discarding an existing working configuration.

Keep the serving window open. Ctrl+C stops serving. The node connects out, so it does not need an inbound port or a public server address.

The app checks each downloaded model description against the catalog before using it. The serving console shows download progress, model loading and warm-up, each request with its token counts, and payments as they are acknowledged and finalized. It uses your saved wallet automatically, and opens a browser only when you choose to.

Run the model campaign

The campaign is an optional diagnostic for a Blackwell card with at least 90 GiB of reported GPU memory. Choose 5. Run model campaign, or run:

$katara = "$env:LOCALAPPDATA\Katara"
$version = (Get-Content "$katara\install.json" -Raw | ConvertFrom-Json).Version
powershell.exe -NoProfile -ExecutionPolicy Bypass -File "$katara\$version\katara.ps1" -Action Campaign

The campaign targets the same Blackwell card with at least 90 GiB of reported GPU memory. It tests these 12 candidates, all at Q4_K_M with an 8,192-token context:

Models	Sizes
Qwen3 Instruct 2507	4B, 30B A3B
Qwen3 Coder Instruct	30B A3B
Qwen3 Next Instruct	80B A3B
Qwen2.5 Instruct	1.5B, 7B, 14B, 32B
Qwen2.5 Coder Instruct	1.5B, 7B, 14B, 32B

Expect about 157 GB of model downloads across the whole campaign. Models run one at a time. Each new test cache is removed after its result is saved; the original 4B smoke-test cache is retained and reused. Peak free-space demand is about 54 GiB. Add -KeepModels to retain the candidate caches if the drive has room for all of them.

Each candidate checks pinned downloads, tokenizer and template hashes, CUDA execution, sampled runtime token IDs, a short speed floor, five repeated synthetic answers in streaming and non-streaming forms, response usage consistency, a prompt near 7,200 tokens, cancellation recovery, and at least two minutes of sustained requests. A separate tool trial tests streaming and non-streaming calls with typed arguments and roughly 200-byte and 2,000-byte documents, at concurrency one. A tool failure is reported separately from the chat result.

Keep the window open. The campaign saves each model’s result, continues after failures, and creates a ZIP under %LOCALAPPDATA%\Katara\data\results. The terminal prints its path. Only synthetic prompts are used; the archive does not include your delegated key. A model can take up to six hours including downloads, with at most 30 minutes for its tool trial.

Rerunning the campaign skips completed results for the same binary, manifest, hardware and recorded driver. To retry a particular candidate, use its name from campaign.json, for example:

powershell.exe -NoProfile -ExecutionPolicy Bypass -File "$katara\$version\katara.ps1" -Action Campaign -Model qwen3-coder-30b -Retest

Concurrency and GPU allocation

Choose 2. Set up wallet, model, GPUs, prices and concurrency after stopping the node. Setup lists the NVIDIA cards and lets you select one card, several cards by number, or all. With two cards, enter 0 to use only GPU 0, 1 to use only GPU 1, or 0,1 to use both. The selection is saved by card identity, so a driver reorder does not change which card runs. Unselected GPUs stay unused by Katara. If a saved card is missing, select the GPUs again in setup.

Selected cards share one active model, split across them by layer. Every selected card must meet the model’s compute requirement and memory per card; combining smaller cards does not bypass it. Up to eight GPUs may be selected.

Every model in the current list serves one request at a time, so setup does not ask for a request limit. The earlier Qwen3 profiles allow 1 to 8 simultaneous requests, chosen in setup.

Before serving, the node checks the selected CUDA devices, allocated slots and context, then measures simultaneous requests at the chosen concurrency. The model must meet its performance floor for every request. If GPU allocation or performance fails, the node stops before advertising capacity; choose a lower concurrency, free GPU memory or change the GPU allocation and retry. A successful startup check validates that configuration on your machine; it does not claim a completed model campaign for other hardware combinations.

Updates and serving activity

The menu ends with 4. Check for an update, 5. Run model campaign, and 6. Exit. Checking shows your installed version and the newest version published for Windows, then asks before downloading the update. Choosing No returns to the menu. Matching versions report No update available. Stop the serving node before installing, then reopen Katara Node from the Start menu. Settings and downloaded models are retained.

Serving occupies its console window. Open http://127.0.0.1:8471/ in a browser on the same computer for live status; there is no separate status menu option. The console itself shows individual requests, token counts, receipt amounts, acknowledgements and finalized payments, plus a summary every minute. See Serving activity for the payment states and saved metadata.

What the checks cover

A local model check establishes behavior for the model, quantization, runtime and hardware tested. Tool support is advertised only where the tool checks passed.

Gateway, receipt and payment acceptance is tested once for the node software integration, rather than repeated for every model.

Campaign records describe the tests that actually ran; they do not claim testing on a 3090, Linux or WSL. Older records contain marketplaceCertified: false because they were created before the release policy changed. There is no need to rerun them to change that field.

The CUDA runtime is pinned to llama.cpp b10941, in a CUDA 12.4 build for compute capability 6.1 to 8.9 and a CUDA 13.3 build for 7.5 and later. If CUDA cannot initialize, the test stops. NVIDIA’s Windows installation guide lists supported Windows versions.

The installed executable is katara-node.exe in a folder named for the installed version, such as %LOCALAPPDATA%\Katara\0.3.020. The app has a console menu and a local serving status page. Keep the serving window open; Ctrl+C stops the node and its GPU runtime. For Linux and the separate WSL path, see Linux and Windows setup.