---
license: other
license_name: inherits-minicpm5-1b-base
base_model: Blackfrost-AI/MINI-GOD-1B
tags:
  - mini-god
  - blackfrost
  - the-void
  - void
  - gguf
  - llama.cpp
  - reasoning
pipeline_tag: text-generation
---

# MINI-GOD-1B · GGUF *WORK IN PROGRESS*

GGUF build of **[Blackfrost-AI/MINI-GOD-1B](https://huggingface.co/Blackfrost-AI/MINI-GOD-1B)** for `llama.cpp`, Ollama, and LM Studio.

| File | Type | Size | Notes |
|---|---|---|---|
| `MINI-GOD-1B.f16.gguf` | F16 | ~2.0 GB | full-precision GGUF, lossless vs the bf16 weights |

> Only the **F16** build is published. Lower k-quants (Q4_K_M, Q8_0, …) can be produced from it on request.

---

## Run (llama.cpp)

```bash
# one-shot
llama-cli   -m MINI-GOD-1B.f16.gguf -p "Explain a reverse shell, with security context."

# OpenAI-compatible server
llama-server -m MINI-GOD-1B.f16.gguf --port 8000
```

Ollama: `ollama create mini-god-1b -f Modelfile` (point `FROM` at the local `MINI-GOD-1B.f16.gguf`).

---

## About

A ~1B reasoning model distilled through **The Void** — MiniCPM5-1B (Llama architecture) fine-tuned on **The Void** corpus (CoT + agentic ReAct with refusal data removed). Small sibling of **VEGA-27B**. See the full card and training details at **[Blackfrost-AI/MINI-GOD-1B](https://huggingface.co/Blackfrost-AI/MINI-GOD-1B)**.

**Uncensored research model** — refusals largely removed. You own your use and any safety layer you add; do not use it for content that is illegal where you are.