Software
Open-weight language, vision and multimodal models you can self-host
Open-weight and openly licensed models make it possible to run language, vision and multimodal inference on your own hardware, subject to each model's licence terms.
Representative platforms with the details engineering and procurement teams shortlist on.
MIT-licensed runtime for pulling and serving open-weight models locally through a CLI and local API.
MIT-licensed C and C++ inference engine for open-weight language models, on CPU or GPU backends.
Apache 2.0 licensed 24-billion-parameter model built for low latency and self-hosted deployment.
MIT-licensed 14-billion-parameter small language model built for low-latency reasoning tasks.
Specifications are summarised from publicly published manufacturer material and are provided for orientation only. Always confirm current figures directly with the manufacturer before purchasing or designing in.
Chip manufacturer
Publishes the Coral Edge TPU accelerators for low-power on-device inference and maintains LiteRT, the on-device runtime that succeeded TensorFlow Lite, together with the open-weight Gemma model family.
Mountain View, California, United States
Software provider
Maintains the YOLO family of open-source vision models covering detection, segmentation, classification, pose estimation and oriented bounding boxes, dual-licensed under AGPL-3.0 and a commercial licence.
Frederick, Maryland, United States
Software provider
Maintains the MIT-licensed Ollama runtime for pulling and serving open-weight language and multimodal models locally on workstations, servers and capable edge systems.
Palo Alto, California, United States
Software provider
Maintainers of the ggml tensor library and llama.cpp, an MIT-licensed C and C++ inference engine that runs open-weight language models on CPUs and on CUDA, Metal and Vulkan backends.
Sofia, Bulgaria
Software provider
Originated ONNX Runtime, the MIT-licensed cross-platform inference accelerator, and publishes the Phi family of small open-weight models aimed at low-latency local inference.
Redmond, Washington, United States
Software provider
Publishes open-weight language and multimodal models, including Apache 2.0 licensed Mistral Small builds designed for latency-sensitive and self-hosted deployments.
Paris, France
Software provider
Hosts open-weight models and publishes the SmolVLM family of compact vision-language models designed explicitly for on-device inference under the Apache 2.0 licence.
New York, United States