The local AI platform, built for everyone, not just developers. Text, voice, image, video and music models run on your own device. Free, private, no setup needed.
The models here are GGUF files of new open models, converted for the runtimes that Hearthscale ships, such as stable-diffusion.cpp. Each one is tested on a consumer graphics card before it is published. Its card names the base model, the runtime build that reads it, the parts that go with it and measured times.
Open Hearthscale Server, go to Explore, and search for the model. The Server downloads the model with the parts that go with it, such as a VAE or a text encoder, and installs the runtime that runs it. Explore models.
You can then use the model in apps such as Image studio, Music studio and Meeting notes, and chat with any chat model in Ash.
Run Hearthscale Server on your most capable device, and every other device with Hearthscale can use the models there. Connections are end-to-end encrypted and work on a local network without internet access. A single device can serve models to an entire organisation.