Global

A unified distributed systems framework bridging centralized model orchestration with decentralized edge intelligence.

A unified distributed systems framework bridging centralized model orchestration with decentralized edge intelligence.

FractalCore is the central server, control plane, and model compilation engine for the Fractal decentralized compute ecosystem.

FractalAndroid is the distributed compute client and edge execution runtime for the Fractal decentralized intelligence framework.

5 GB of RAM. Weights stream from SSD; a trained router predicts loads a step ahead. Mac, iPhone, Android, Windows supported

Open-source State-of-the-art TTS model which runs on a CPU 😻

MiniCPM-V and MiniCPM-o are multimodal LLM series designed for strong performance and efficient deployment on devices. MiniCPM-V focuses on efficient vision-language understanding across image, video and text inputs.

FreeToken brings datacenter-scale model serving to your desktop. Run massive models locally, fast and efficiently.

An open-source AI accelerator, developed by AI: RTL, ISA, simulator, compiler and profiler in one repo. 5 on a Kintex-7 PCIe card.

Speech-to-text, text-to-speech, speaker diarization, speech enhancement, source separation, and VAD using next-gen Kaldi with onnxruntime without Internet connection.

cpp — a real-time multilingual ASR inference engine on CPU using BitNet I2S quantization, achieving RTF Code ] [ Models ] [ Report ]

The world is rushing to build AI systems. It is not engineering them.

Automation foundation model for tiny devices: 2-bit, 8-29 MB, tool calls, ASR, structured extraction and embeddings on phones, wearables, smart homes, robots, cars and microcontrollers.

The GIF is an original-speed render of a real local Snake run. Every move calls Laya; the visible cycle safety layer can correct unsafe proposals. The latency figures above are the separate one-question API benchmark, not the frame time of the three-question Snake loop.

Qdrant - High-performance, massive-scale Vector Database and Vector Search Engine for the next generation of AI. Also available in the cloud

A unified library of SOTA model optimization techniques like quantization, distillation, pruning, neural architecture search, speculative decoding, etc. It compresses deep learning models for downstream deployment frameworks like TensorRT-LLM, TensorRT, vLLM, etc.

A high-performance inference engine for AI models. It allows you to deploy AI directly in your app with zero latency, full data privacy, and no inference costs. Key features:

We are releasing MiniCPM5-2B, the second model in the MiniCPM5 series, following MiniCPM5-1B. It is a dense 2B Transformer that scales up the same training recipe, built for on-device, local deployment, and resource-constrained scenarios, reaching 2B-class open-source SOTA.

Cross-platform, customizable ML solutions for live and streaming media.

MNN: A blazing-fast, lightweight inference engine battle-tested by Alibaba, powering high-performance on-device LLMs and Edge AI.

Moonshine Voice is an open source AI toolkit for developers building real-time voice agents and applications.

System requirements to run python examples on x86-64 systems:

jevos: an open-source alternative to Jev and Microsoft-Decision-1. Yes/no decisions, choices and scores with probabilities, on your laptop's CPU.

3 GB model into memory. 35 GB core and FP16 KV cache in memory, then streams only the experts needed for each token from SSD. This is what lets the model run on Macs with 8 GB of RAM.

9 million parameter language model that generates text on an ESP32-S3 microcontroller. 88 tokens per second on a small screen wired to the chip. It fits because most of the model lives in flash instead of RAM, using Per-Layer Embeddings, an idea from Google's Gemma 3n.

Your iPhone helps your Mac run a 27B model: faster prompt reading and more context over a USB-C cable

ncnn is a high-performance neural network inference framework optimized for the mobile platform

io, the cloud and edge native messaging system.

Eight SDKs, one C++ core. A capability registry routes every call to the best engine on the device, and the Console manages your fleet from above.

AI and the OpenMOSS team. 1B parameters, it is designed for realtime speech generation, can run directly on CPU without a GPU, and keeps the deployment stack simple enough for local demos, web serving, and lightweight product integration.