From: <huggingface.co/mlx-community> Newsgroups: alt.freeweights.qwenSubject: Qwen3.8-27B-MTP-4bit Date: 2026-11-14Commit: 745ac04Signed: 2DC28CC7B26A283B ✓ Text generation model with 425M parameters. Safetensors weights, 258 MB in all. Published by mlx-community under Apache-2.0.
425M params 258 MB Qwen3.8-27B-MTP-4bit safetensors permissive Facts Params 425M Size 258 MB Formats safetensors Task text generation License Apache-2.0 permissive
fw
$ fw get mlx-community/Qwen3.8-27B-MTP-4bit
# manifest sha256 7424393efdee… · 258 MB · 9 files
$ fw verify ./Qwen3.8-27B-MTP-4bit mlx-community/Qwen3.8-27B-MTP-4bitHosted (signed) mlx-community/Qwen3.8-27B-MTP-4bit Last check: 2026-11-14 21:30 UTC
Get help: custom fine-tune or private deploy. We review your setup before we quote.
Quantized of Qwen3.8-27B .
Works with
Runs with Ranked by fit to this model (safetensors), then by GitHub stars.
Show all 24 Fine-tune with Show all 29
Overview
Size 258 MB
Weights 239 MB (safetensors)
Parameters 425M
Formats safetensors
Task text generation
License Apache-2.0 (permissive)
Files 9, each with a sha256
Publisher mlx-community
Added 2026-11-14
Section alt.freeweights.qwen Lineage
Original release from the publisher.
Base model: Qwen/Qwen3.8-27B
Where these bytes come from
Publisher at the pinned commit →You
Sources are tried in order. A file counts only when its size and sha256 match the manifest. Sources are checked when you ask for them, not on a timer.
Sources
Badge Host Last check Result huggingface.co 2026-11-14 21:30 UTC passed
Model card From the publisher at commit
745ac04 . Shown as plain text with links; images open at the publisher.
Qwen3.8-27B-MTP-4bit#
This repository contains the Multi-Token Prediction (MTP) drafter weights split from Qwen/Qwen3.8-27Bmlx_vlm.convert for mlx-vlm speculative decoding.
This is not a standalone model. Use it as the draft model alongside a compatible Qwen3.8 27B target checkpoint.
Use with mlx-vlm#
mlx_vlm generate \
--model mlx-community/Qwen3.8-27B-4bit \
--draft-model mlx-community/Qwen3.8-27B-MTP-4bit \
--prompt "Write a quicksort in Python." \
--max-tokens 256 \
--enable-thinking--draft-kind mtp is detected automatically from model_type.
Model details#
Model type: qwen3_5_mtp MTP block size: 3 Target architecture: Qwen3.8 27B Precision: MLX affine 4-bit, group size 64 Runtime: MLX / mlx-vlm Source revision: 1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0 The adapter holds only the MTP drafter weights; the target supplies the token embeddings and language-model head at runtime. Use an adapter and target derived from the same Qwen3.8 27B checkpoint. The upstream Apache 2.0 license and model limitations continue to apply.
Files 9 files · 258 MB Weights 239 MB (safetensors). Sizes and sha256 from the signed manifest.
Hide all 9 files with their sha256 File Size sha256 .gitattributes 1.57 KB 34448b82c17d60fec9b65b1f093c115ddbaadc04beb1b0140b6bfed2e012a930 README.md 1.42 KB 61b0a92d432cd2c3095a86bbe04eb4e11671f96bc506361cce4006277b6568ca chat_template.jinja 8.95 KB c3cf9e34abf4f9e36c2d72165aa9c132d3e2a725b6c2586aaa3a8af9d7a81041 config.json 3.80 KB 16094efa6177985ab3725a9d6d61d6ab248b71e4b42a114efddcdb2aaddc0a55 model.safetensors 239 MB 76663c101e7e8ea9c0ae17bcb95183cd7f733ce424c912b8b264a7b1c48e4cc6 model.safetensors.index.json 1.97 KB 6233c1805e07b028f4eb34d50dac5bf22a676728eeaf2ea43c4f94523af7a4fc tokenizer.json 12.8 MB 0997f410c57a1f4e53b09e4be8f4a172d90edd9564368fb0847030937229b9f3 tokenizer_config.json 1.12 KB 66e427c470fe580fe8c7b5725d857af23d8417e37fae62667ec698306a19987b vocab.json 6.72 MB ce99b4cb2983d118806ce0a8b777a35b093e2000a503ebde25853284c9dfa003
Versions Each time the pinned upstream commit or file list changed in this repo. Newest first.
Upstream commit Date Contents 745ac04da2 pinned now manifest updated, same commit 2026-11-14 21:30 UTC 9 files · 258 MB 745ac04da2 2026-11-14 21:30 UTC 9 files · 258 MB
Use it fw
$ fw get mlx-community/Qwen3.8-27B-MTP-4bit
# manifest sha256 7424393efdee… · 258 MB · 9 files
$ fw verify ./Qwen3.8-27B-MTP-4bit mlx-community/Qwen3.8-27B-MTP-4bitfw.py Checks the signature and every sha256, tries each source in order, resumes.
$ python3 fw.py get https://freeweights.allternit.com/manifests/mlx-community/Qwen3.8-27B-MTP-4bit.json --dir ./Qwen3.8-27B-MTP-4bit
$ python3 fw.py verify ./Qwen3.8-27B-MTP-4bit ./mlx-community__Qwen3.8-27B-MTP-4bit.jsonhf The Hugging Face CLI with one setting. Same repo id.
$ HF_ENDPOINT =https://freeweights.allternit.com hf download mlx-community/Qwen3.8-27B-MTP-4bit --revision 745ac04da2f19acd24e1f23b88385e3aade93129 --local-dir ./Qwen3.8-27B-MTP-4bittransformers Load the weights in Python.
import os
os.environ["HF_ENDPOINT" ] = "https://freeweights.allternit.com"
from transformers import AutoModelForCausalLM, AutoTokenizer
tok = AutoTokenizer.from_pretrained("mlx-community/Qwen3.8-27B-MTP-4bit", revision="745ac04da2f19acd24e1f23b88385e3aade93129")
model = AutoModelForCausalLM.from_pretrained("mlx-community/Qwen3.8-27B-MTP-4bit", revision="745ac04da2f19acd24e1f23b88385e3aade93129", device_map="auto")vLLM Serve an OpenAI-compatible API.
$ HF_ENDPOINT =https://freeweights.allternit.com vllm serve mlx-community/Qwen3.8-27B-MTP-4bit --revision 745ac04da2f19acd24e1f23b88385e3aade93129License Name Apache License 2.0 Id Apache-2.0 Class Permissive Redistribution yes Listing Hosted (signed) Decided License as declared by the publisher. No person read it. 2026-11-14 Text in README.md Text sha256 61b0a92d432cd2c3095a86bbe04eb4e11671f96bc506361cce4006277b6568ca Terms gate No extra terms beyond the license. Read the license text
Licenses are checked against the license table . A custom license passes only after a person reads it.
Provenance From: <huggingface.co/mlx-community> Newsgroups: alt.freeweights.qwenSubject: Qwen3.8-27B-MTP-4bit Date: 2026-11-14Commit: 745ac04Signed: 2DC28CC7B26A283B ✓ Upstream mlx-community/Qwen3.8-27B-MTP-4bit Pinned commit 745ac04da2f19acd24e1f23b88385e3aade93129 Manifest Qwen3.8-27B-MTP-4bit.json · .minisig Manifest sha256 7424393efdeeed65c797fcad32b5bad484e875d94edba087cf0336f66e9c67a9 Signature minisign, Ed25519, key 2DC28CC7B26A283B Created 2026-11-14 21:30 UTC Generator freeweights-manifest/0.1.0 Takedowns None. How to verify
$ curl -O https://freeweights.allternit.com/manifests/mlx-community/Qwen3.8-27B-MTP-4bit.json
$ curl -O https://freeweights.allternit.com/manifests/mlx-community/Qwen3.8-27B-MTP-4bit.json.minisig
$ minisign -Vm Qwen3.8-27B-MTP-4bit.json -P RWQ7KGqyx4zCLXtUFOqvfv8L3g+8fpKJ1FsuxPyOYMF+4KufbUxKq1Oq
$ shasum -a 256 <file> # compare with the Files tab A manifest whose signature fails must be rejected. A file whose size or sha256 does not match must be discarded. Verify a download
Model tree No model in the catalog names this one as its parent.
Get it
$ HF_ENDPOINT =https://freeweights.allternit.com hf download mlx-community/Qwen3.8-27B-MTP-4bitOpen in Freeweights Studio Pull, verify and run it in the Allternit app on Desktop, web or phone.
2.2B params 4.37 GB Aquila-VL-2B-llava-qwen safetensors permissive Facts Params 2.2B Size 4.37 GB Formats safetensors Task visual question answering License Apache-2.0 permissive flip From: <huggingface.co/BAAI> Subject: Aquila-VL-2B-llava-qwen 34.7B params 69.3 GB Aurora-Code-1 safetensors permissive Facts Params 34.7B Size 69.3 GB Formats safetensors Task text generation License Apache-2.0 permissive flip From: Perciqa <huggingface.co/Perciqa> Subject: Aurora-Code-1 7.6B params 30.6 GB bge-reasoner-embed-qwen3-8b-0923 safetensors permissive Facts Params 7.6B Size 30.6 GB Formats safetensors Task feature extraction License Apache-2.0 permissive flip From: <huggingface.co/BAAI> Subject: bge-reasoner-embed-qwen3-8b-0923 9.3B params 18.5 GB BMOJOF-primed-HQwen3-8B-Instruct safetensors permissive Facts Params 9.3B Size 18.5 GB Formats safetensors Task text generation License Apache-2.0 permissive flip From: amazon <huggingface.co/amazon> Subject: BMOJOF-primed-HQwen3-8B-Instruct qwen 658 MB Bubba-Qwen3.6-27B-LoRA safetensors permissive Facts Params not stated Size 658 MB Formats safetensors Task not tagged License Apache-2.0 permissive flip From: nbeerbower <huggingface.co/nbeerbower> Subject: Bubba-Qwen3.6-27B-LoRA qwen 447 GB bytkim_Qwen3.8-27B-pi-GGUF gguf permissive Facts Params not stated Size 447 GB Formats gguf Task image text to text License Apache-2.0 permissive flip From: bartowski <huggingface.co/bartowski> Subject: bytkim_Qwen3.8-27B-pi-GGUF
mlx-community <huggingface.co/mlx-community>Newsgroups: alt.freeweights.qwenSubject: 
BAAI