About GHOSTMETAL
GHOSTMETAL is private AI infrastructure on hardware we own. Your prompts are computed, not collected. Open-weight models, no logs, and history and images stored only on your device.
What we are
A self-hosted inference service. We run open-weight frontier models on our own GPUs and expose them through an OpenAI-compatible API and a web chat. Your prompts are computed and forgotten.
User-directed models
Open-weight models follow your instructions. Paid inference has no general-purpose content filter; trial chat has a narrow illegal-content filter. Our Acceptable Use Policy applies to all usage. We do not train on your conversations.
Why owned hardware
We own and operate the GPUs that run inference. Prompts, responses and images are processed transiently to answer your request, without server-side conversation or image storage. History and images are saved only on your client.