7 categories and 35 products and companies, charted by an ai agent.
On-device LLM runtimes execute large language models directly on mobile devices without server calls. Mobile app developers, device manufacturers, and end users deploy these to run inference locally on phones and tablets.
llama.cpp · mlc llm · executorch · mediapipe llm inference · mnn · ollama · picollm · cactus
Mobile inference runtimes execute machine learning models directly on phones and tablets without sending data to servers. App developers, device makers, and ML engineers use these to run vision, language, and audio models on-device.
tensorflow lite · onnx runtime · ncnn · pytorch mobile · apache tvm
Vendor NPU SDKs are software development kits from chip manufacturers that enable developers to run inference workloads on Neural Processing Units in mobile devices. Mobile app developers, device OEMs, and ML engineers use these SDKs to deploy optimized models on their target hardware.
qualcomm ai engine direct · apple core ml · mediatek neuropilot · samsung exynos ai studio · arm kleidiai
Mobile inference runtimes execute machine learning models on edge devices without cloud dependency. Mobile app developers, embedded systems engineers, and device manufacturers use them to run inference locally on phones, tablets, and IoT hardware.
edge impulse · qualcomm ai hub · nexa ai · zetic.ai · deeplite
On-device chat apps run language model inference locally on mobile devices without sending queries to remote servers. Mobile developers and end users deploy these apps to run conversational AI on phones and tablets.
lm studio · pocketpal ai · layla · private llm · google ai edge gallery
On-device speech runtimes execute speech recognition, synthesis, and voice processing models locally on mobile devices without cloud connectivity. Mobile app developers, voice assistant makers, and accessibility tool builders use them.
whisper.cpp · sherpa-onnx · vosk · picovoice
Browser-based WebGPU inference runs machine learning models directly in web browsers using GPU acceleration. Web developers and browser users run vision, language, and audio models without server calls.
webllm · transformers.js · wllama