Open source. Open possibilities.

WhisperLiveKit is a self-hosted, ultra-low-latency speech-to-text pipeline. It provides real-time transcription, speaker diarization, and translation via WebSocket and OpenAI/Deepgram-compatible APIs.

jetson-inference is a C++ and Python deep-vision library for NVIDIA Jetson. It uses TensorRT for on-device inference and PyTorch for training, with examples, pre-trained models, and tutorials for classification, detection, segmentation, pose, action, depth, and camera streaming.

MediaPipe is Google's cross-platform framework and suite of on-device machine learning solutions for live and streaming media, covering vision, text and audio tasks on Android, iOS, web, desktop and edge devices.

SANA is NVIDIA's open-source family of efficient diffusion models for high-resolution image and video generation, with training and inference code, linear attention, DC-AE compression and 4-bit deployment.

YOLOv5 is a fast, accurate computer vision framework in PyTorch for object detection, instance segmentation, and image classification. It supports training on custom datasets, inference from various sources, and exporting models to formats like ONNX, TensorRT, and CoreML.

ferrite is a Rust-native inference engine for GLM-5.3-Flash, featuring PDAF disaggregation, composable TP/CP/DCP parallelism, and exact MHC. It supports GatedDeltaNet and DSA sparse attention with CUDA kernels for sm_100a.

A local multimodal LLM plugin for Unreal Engine 5 providing offline inference for text, speech-to-text, and text-to-speech.

The official Python SDK for the Model Context Protocol (MCP), enabling the creation of standardized servers and clients for LLM applications.

TextGen is an open-source desktop app for running local LLMs with text, vision, tool-calling and web search. It offers an OpenAI/Anthropic-compatible API, multiple backends, LoRA training and image generation, all fully offline with no telemetry.

XGBoost is an optimized distributed gradient boosting library that is efficient, flexible, and portable. It implements machine learning algorithms under the Gradient Boosting framework and provides parallel tree boosting (GBDT, GBM) to solve data science problems accurately and quickly. It runs on major distributed environments and supports multiple languages.

Strata is a native machine learning and linear algebra library written in Mojo, offering scikit-learn-compatible estimators, dense/sparse matrix containers, zero-copy serialization, and composable pipelines with SIMD and LAPACK acceleration.

Spring AI is a Spring-native application framework for AI engineering, offering portable, strongly-typed APIs that connect enterprise data and APIs with major AI model providers and vector stores.

FastChat is an open platform for training, serving, and evaluating large language model chatbots, powering Chatbot Arena. It offers distributed multi-model serving with a web UI and OpenAI-compatible APIs, plus Vicuna and MT-Bench.

vLLM Ascend is a community-maintained hardware plugin enabling vLLM to run seamlessly on Ascend NPUs, supporting Transformer, MoE, embedding, and multimodal LLMs.

Batchalign3 is a TalkBank audio and ML pipeline for producing and enriching CHAT transcripts, covering ASR, forced alignment, neural morphotagging, translation and utterance segmentation. It ships a CLI, Python package, PyO3 bridge, React dashboard and an experimental Tauri desktop shell.

A Python library providing a unified high-level API for various local Large Language Model (LLM) inference backends, starting with llama-cpp-python.

ComfyUI is a modular AI creation engine featuring a visual node graph interface for generating images, videos, audio, and 3D models using state-of-the-art open-source and closed-source models.

Modular Platform is an open-source unified platform for AI development and deployment, hosting the MAX framework and Mojo language, including compiler, standard library, accelerator kernels, and an OpenAI-compatible inference server.

PyScript is an open source platform that runs Python in the browser using Pyodide (CPython) and MicroPython compiled to WebAssembly. Add a stylesheet and script tag, then write Python in HTML script tags to build rich browser apps on any device.

PaddlePaddle is an industrial-grade deep learning platform providing a comprehensive framework for machine learning, including model libraries and development kits.