Hugging Face released Humanizer, a 12B model that rewrites AI drafts to sound human. It preserves every fact, number, date, name and quote. It runs locally on a user’s machine. No AI detector was used in its training. It supports English and Chinese. Multiple quantized versions are available, down to 3.9 GB for 8 GB machines. A desktop app for macOS and Windows auto-selects a model size. A command-line tool handles long documents and agents. GGUF files now include built-in sampling settings and a chat template. Quantization-aware training improved smaller models’ accuracy. The 2-bit version scores lowest on AI detectors but makes more fact errors.
What changed
AI-written text often retained detectable patterns and required manual editing to sound human.
What it unlocks
Rewriting AI-generated drafts to sound human-like while preserving all facts, numbers, and quotes.
- 12B parameter model
- 8 GB minimum memory for smallest quantized version
- 23.8 GB for full precision (bf16)
- 0.0017 KL divergence for Q8_0 quantized version
What you need to act on it
- local execution with minimum 8 GB RAM
- compatible runtime (llama.cpp, MLX, transformers, vLLM, Ollama, LM Studio)
Sources