humanizer
A 12B model that rewrites AI-written drafts (emails, essays, reports, forum posts; English and Chinese) so they read like a person wrote them. It is trained to keep every number, unit, date, name and quote and to add nothing, and it runs locally.
Params
12.0 B
Context
—
Downloads 30d
139 K
Likes
2,139
Download history
tracking started — chart appears after 7 days of snapshots (1 recorded)139 K downloads in the last 30 days
Can you run it?
Estimated VRAM at 8K context unless noted. Pick your hardware to see the verdict per quantization.
| File | Quant | Size | Est. VRAM | Verdict on RTX 4090 · 24 GB |
|---|---|---|---|---|
| humanizer-12b-IQ2_XS-QAT.gguf | IQ2_XS | 3.9 GB | 6.6 GB | ✅ Runs comfortably |
| humanizer-12b-Q3-QAT.gguf | Q3 | 5.6 GB | 8.4 GB | ✅ Runs comfortably |
| humanizer-12b-Q4_K_M.gguf | Q4_K_M | 7.6 GB | 10.7 GB | ✅ Runs comfortably |
| humanizer-12b-Q6_K.gguf | Q6_K | 10.0 GB | 13.3 GB | ✅ Runs comfortably |
| humanizer-12b-Q8_0.gguf | Q8_0 | 12.7 GB | 16.2 GB | ✅ Runs comfortably |
| humanizer-12b-bf16.gguf | GGUF | 23.8 GB | 28.5 GB | ❌ Won’t fit |
| model.safetensors | bf16 | 23.9 GB | 28.6 GB | ❌ Won’t fit |
Estimate: file size × 1.1 + KV cache at 8K + 0.5 GB overhead. Not a benchmark — how we calculate this.
Run it
copy-paste, exact tags checked against the Hub$ ollama run humanizer # pin the quantization explicitly $ ollama run humanizer-q4_k_m
$ huggingface-cli download jialinyyzz/humanizer-GGUF \
humanizer-12b-Q4_K_M.gguf --local-dir .
$ llama-cli -m humanizer-12b-Q4_K_M.gguf \
-c 8192 -ngl 99 -t 8 --color
$ curl -s https://aimodelscomparison.com/api/v1/models/humanizer
{
"hf_id": "jialinyyzz/humanizer",
"params_b": 11.96,
"context_length": null,
"license": { "id": "apache-2.0", "commercial": "yes" },
"downloads_30d": 138556,
"vram_estimates": [
{ "quant": "IQ2_XS", "gb": 6.6 },
{ "quant": "Q3", "gb": 8.4 }
],
"updated_at": "2026-10-25T01:00:36Z"
}
Specifications
- Architecture
- Gemma4UnifiedForConditionalGeneration
- Parameters
- 12.0 B
- Tensor type
- BF16
- Licence
- apache-2.0
- First seen on the Hub
- 2026-09-11
- Training datasets
- undisclosed
- Added to our catalog
- 2026-10-25
Compare with any text-generation model