llmdot
Run local GGUF language models from .NET — one package, one format, one programming model. Native in-process inference for C# and the whole .NET ecosystem, no Python and no external server.
Key Features
One NuGet Package
Add Llmdot to any .NET project and run GGUF models in-process. No sidecar, no HTTP, no Python.
Native GGUF Inference
Load and run quantized GGUF models directly from C#, F#, or VB.NET through a single programming model.
Microsoft.Extensions.AI
Plugs into the standard .NET AI abstractions so it drops into existing apps and pipelines.
Cross-Platform
Runs anywhere .NET runs — Windows, Linux, and macOS — with the same code and the same model files.
Local & Private
All inference stays on your hardware. No cloud calls, no per-token billing, works fully offline.
Familiar to .NET Devs
Idiomatic C# API — no FFI boilerplate, no marshalling ceremony. Just install and call.
Quick Start
dotnet add package Llmdot.Coreusing Llmdot;
// Load a local GGUF model
using var model = LlmModel.Load("llama-3.2-1b-instruct.Q4_K_M.gguf");
// Generate text in-process — no server, no Python
var response = model.Generate("Explain gravity in one sentence.");
Console.WriteLine(response);Why llmdot
The .NET ecosystem has lacked a first-class way to run local LLMs. Existing options mean shelling out to an external server, wrapping Python, or hand-rolling P/Invoke against llama.cpp. llmdot gives .NET developers the same in-process, native experience that Python and Rust developers already enjoy in the rest of the Cognisoc stack: one package, one model format, one programming model.
- In-process — the model runs inside your application. No HTTP hop, no daemon to manage.
- GGUF-native — use the same quantized model files as the rest of the local-LLM world.
- Standard abstractions — implements
Microsoft.Extensions.AIinterfaces so it composes with the tools you already use.
Where it fits
llmdot is the .NET member of the Cognisoc family of local inference runtimes — the same philosophy as mullama (Python/Go/PHP/Node/Rust) and llamafu (mobile/Flutter), built for the languages .NET teams actually ship in.