Short version: deepgrav-30b, the stock model for DeepGravity, is published on Hugging Face, free under the Apache License 2.0.
It is a single 30B GGUF file, Q5_K_M quantization, about 19.8 GB, with the 12 Principles baked into its chat template — the Principles are what make an engine a DeepGravity engine. It runs in llama.cpp or Ollama, the same as any other GGUF of its family. The model card on Hugging Face has the full details: what it is based on, the file hashes, and the template.
By Dora
Leave a Reply