Software
Compilers, quantisation, profiling and deployment SDKs
Developer tools turn a trained model into something that runs efficiently on target silicon: graph compilers, quantisation and pruning tools, profilers and device SDKs.
Representative platforms with the details engineering and procurement teams shortlist on.
Official Jetson software stack combining the Linux board support package with CUDA-accelerated AI libraries.
Inference compiler and runtime family that optimises models for NVIDIA GPUs, including Jetson.
Apache 2.0 licensed toolkit for optimising and deploying inference across Intel CPUs, GPUs and NPUs.
MIT-licensed cross-platform inference accelerator with hardware-specific execution providers.
Compiler that maps trained models onto Hailo accelerators, part of the Hailo AI Software Suite.
Edge MLOps platform for building and deploying models to microcontrollers, NPUs, CPUs and GPUs.
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
Develops the Jetson family of edge system-on-modules, the DGX Spark desktop AI system and the JetPack, TensorRT and DeepStream software stacks used for edge inference, robotics and vision AI.
Santa Clara, California, United States
Chip manufacturer
Supplies Core Ultra client processors that combine CPU, Arc graphics and an integrated NPU, and maintains the open-source OpenVINO toolkit for optimising inference on Intel CPUs, GPUs and NPUs.
Santa Clara, California, United States
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
Edge AI startup
Fabless company developing discrete edge AI processors, M.2 and PCIe accelerator modules and the Hailo AI Software Suite, spanning classic vision workloads and, with Hailo-10H, generative models.
Tel Aviv, Israel
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
Provides computer vision tooling including the open-source Inference package and inference server, covering model loading, pre- and post-processing and workflow execution on edge devices.
Des Moines, Iowa, United States
Software provider
Edge MLOps platform, now part of Qualcomm, for collecting data, training and deploying models to microcontrollers, NPUs, CPUs and GPUs across a broad partner hardware ecosystem.
San Jose, California, United States
Software provider
Container-based platform for deploying and managing fleets of Linux edge devices, with over-the-air updates, an API and SDK, and support for a large catalogue of device types.
Remote / Seattle, 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
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