Qwen3.8-27B-NVFP4-MTP-GGUF
A family of nine GGUF files of Qwen3.8-27B (the native vision-language 27B dense model, Gated DeltaNet + Gated Attention hybrid layout, 262,144-token native context, MTP speculative head), converted from unsloth/Qwen3.8-27B-NVFP4. The MTP (multi-token prediction) speculative head is baked into every file — no separate drafter needed.
Params
27.0 B
Context
—
Downloads 30d
275 K
Likes
121
Download history
daily snapshots · 65 days
▲ 466 K in the last 30 days (62.9%)
735 K263 K
Oct 12Oct 22Nov 1Nov 10
769 K193 K
Sep 7Sep 28Oct 20Nov 10
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 |
|---|---|---|---|---|
| Qwen3.8-27B-NVFP4-MTP-ORIG.gguf | GGUF | 33.1 GB | 41.0 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 qwen3-8-27b-nvfp4-mtp-gguf # pin the quantization explicitly $ ollama run qwen3-8-27b-nvfp4-mtp-gguf-gguf
$ huggingface-cli download esatapedico/Qwen3.8-27B-NVFP4-MTP-GGUF-GGUF \
Qwen3.8-27B-NVFP4-MTP-ORIG.gguf --local-dir .
$ llama-cli -m Qwen3.8-27B-NVFP4-MTP-ORIG.gguf \
-c 8192 -ngl 99 -t 8 --color
$ curl -s https://aimodelscomparison.com/api/v1/models/qwen3-8-27b-nvfp4-mtp-gguf
{
"hf_id": "esatapedico/Qwen3.8-27B-NVFP4-MTP-GGUF",
"params_b": 27.00,
"context_length": null,
"license": { "id": "apache-2.0", "commercial": "yes" },
"downloads_30d": 274990,
"vram_estimates": [
{ "quant": "GGUF", "gb": 41.0 }
],
"updated_at": "2026-10-17T01:00:09Z"
}
Specifications
- Parameters
- 27.0 B
- Licence
- apache-2.0
- First seen on the Hub
- 2026-08-14
- Training datasets
- undisclosed
- Added to our catalog
- 2026-09-07
Compare with any text-generation model