TensorSharp
A native .NET inference engine for GGUF models that runs autoregressive and text-diffusion large language models, image generation and editing, and video generation, with a console app, browser chat UI and OpenAI/Ollama-compatible HTTP APIs.
TensorSharp runs autoregressive LLMs and DiffusionGemma-style text-diffusion models, Qwen-Image-2.1 generation and editing, and MiniMax-H3 video with native 32 kHz stereo audio, plus Wan 2.1 and 2.2 for video alone, entirely in managed .NET code with native CPU and accelerator backends. Supported model families include DeepSeek V4, GLM 5.x, Gemma 4, several Qwen variants, GPT-OSS, Nemotron-H, Mistral 3 and Hunyuan Dense, alongside BERT and XLM-R text and code embedding models.
An optional TensorSharp.AgentHost layer adds Agent Skills, a bounded, in-process model-to-tool loop for sandboxed file and shell work, and bounded automatic subagent delegation. Served endpoints are Ollama- and OpenAI-compatible, alongside a console application and a browser-based chat interface.
It runs on Windows, macOS, iOS and Linux with GPU acceleration, and is written in C#.