CatalogWantedMirrorsLedgerOperating StatementDocsStudio statusCard system
alt.freeweights.qwen / Qwen3.8-Flash-Next-4bit
From: mlx-community <huggingface.co/mlx-community>Newsgroups: alt.freeweights.qwenSubject:

Qwen3.8-Flash-Next-4bit

Date: 2026-11-14Commit: 07b5dc6Signed: 2DC28CC7B26A283B ✓

Image text to text model with 177.4B parameters. Safetensors weights, 112 GB in all. Published by mlx-community under Qwen-Community-1.0 (custom).

177.4B params112GBQwen3.8-Flash-Next-4bitsafetensorscustom terms
FactsParams177.4BSize112 GBFormatssafetensorsTaskimage text to textLicenseQwen-Community-1.0 (custom)From sourcecustom termsread terms
# custom terms: read the license before you download
$ fw get mlx-community/Qwen3.8-Flash-Next-4bit
# manifest sha256 de1ed434dcb6… · 112 GB · 34 files
$ fw verify ./Qwen3.8-Flash-Next-4bit mlx-community/Qwen3.8-Flash-Next-4bit
From sourceHosted (signed)mlx-community/Qwen3.8-Flash-Next-4bitLast check: 2026-11-14 21:30 UTC
Get help

Get help: custom fine-tune or private deploy. We review your setup before we quote.

Quantized of Qwen3.8-Flash-Next.

Works with

Run23Fine-tune29Convert15Evaluate13Scan13Apps62
OverviewModel cardFilesUse itLicenseProvenanceModel tree

Overview

Size
112 GB
Weights
112 GB (safetensors)
Parameters
177.4B
Formats
safetensors
Task
image text to text
License
Qwen-Community-1.0 (custom) (custom)
Files
34, 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-Flash-Next

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
BadgeHostLast checkResult
From sourcehuggingface.co2026-11-14 21:30 UTCpassed

Model card

From the publisher at commit 07b5dc6. Shown as plain text with links; images open at the publisher.

Qwen3.8-Flash-Next-4bit (MLX)#

4-bit MLX conversion of Qwen/Qwen3.8-Flash-Next, converted with mlx-vlm main (post-#2032) at commit d1bd74ed, group size 32.

Group size 32 is required so the PLE n-gram embedding dimensions can be quantized.

Usage#

pip install git+https://github.com/Blaizzy/mlx-vlm
mlx_vlm.generate --model mlx-community/Qwen3.8-Flash-Next-4bit \
  --prompt "Explain sparse attention in one paragraph." --max-tokens 256

Why this conversion exists#

Qwen4ExpRMSNorm applies 1 + w to norm gains that the released checkpoint stores centered at zero, matching upstream Qwen4ExpTextRMSNorm. Several MLX conversions published before #2032 landed were made with a converter that folded that +1 into the saved weights, so loading them applies the offset twice and generation degenerates into noise (#2041).

Verification#

Checked against the bf16 source after conversion:

checkresult
source integrity131/131 shards, tensor bytes byte-exact vs index total_size
norm gain center (source vs converted)+0.2216 vs +0.2216, delta +0.00000
norm tensors bit-identical148 / 148
sanitize() idempotencepasses over 480 1-D tensors
generationcoherent output at --temperature 0.0

A gain center near +1.15 instead of +0.22 is the signature of the double-shifted conversions described above.

Files

34 files · 112 GBWeights 112 GB (safetensors). Sizes and sha256 from the signed manifest.
Show all 34 files with their sha256
FileSizesha256
.gitattributes1.57 KB34448b82c17d60fec9b65b1f093c115ddbaadc04beb1b0140b6bfed2e012a930
README.md1.67 KBeb54592ce293006d75d181be8dca197282f3b76148852c9c505fc9242bcef3fe
chat_template.jinja8.95 KBc3cf9e34abf4f9e36c2d72165aa9c132d3e2a725b6c2586aaa3a8af9d7a81041
config.json80.8 KB8da41ff81dd216be279dfd4070d3eb2b9b84517c3f892351b32e5680e6e1d6ce
generation_config.json202 Be70c136c1b78ddc1fb0905bac8e733a4dc448d4f852a5dd75143fffc70be550e
model-00001-of-00022.safetensors5.28 GB879e2028f98cc983b750ed876a005f1ad125e0cb61fb4f49c93695350eefe45c
model-00002-of-00022.safetensors5.25 GBe4ac7254b17359f728bda19da413278c399b14b5ef849817febc73a223894060
model-00003-of-00022.safetensors5.25 GBbed34392682d6431b612af18438aa7f43091cfa29f71cd78133211302676544d
model-00004-of-00022.safetensors5.25 GB7dc7d7f0887777814665f9b0674788dced9d7263eb42c12e9f44d4cb823c5cf2
model-00005-of-00022.safetensors5.25 GBac100ce27255aba1f11b09f2ee9b35edd108a307f5699a9dafee878668688abd
model-00006-of-00022.safetensors5.25 GB7b330de2e041d8cee9a03b2b58e4e34b22983692f1e702c19caf58d09ad25aa0
model-00007-of-00022.safetensors5.07 GB51dbb8e92a5be5ad79a8a3ddf7da8f1696a01e093de0045c45e7b67a4d6a355f
model-00008-of-00022.safetensors5.33 GBeece6ac6c12a46c817f0eb4e334437c80eb8d4106bea7618c7b44c40662d5dfb
model-00009-of-00022.safetensors5.33 GBb0f449ff11c75bc0fc092ef306c8b6fdfe07b401d93b52b1557302cc1c9a1b15
model-00010-of-00022.safetensors4.97 GB6b9a74feb7a6aaeaac0d886b6dae312bfdc96eae7122f5b6f923cb9a1ed607f2
model-00011-of-00022.safetensors5.34 GB03bb7d1be20bea990ce2cade373a3725b180a181cbc7f5549d9ba467979c4144
model-00012-of-00022.safetensors4.96 GB21c581028c4ae75a97cd5dadc37fc578f495c6a8229b5f25d35fd7f32b2efea4
model-00013-of-00022.safetensors5.33 GBf14e42d49a1421cd4c1b37719771a55a9c979789d9159ebc8f73a901b196df43
model-00014-of-00022.safetensors5.33 GBc198a9ba1c99f91e54330e34f94ab758c1f2e9f246585f0eedba0f416fd066d1
model-00015-of-00022.safetensors4.97 GB4991656e138e6c80015e0b4b31acd628c97601522c59f3fe94121e991efa74fa
model-00016-of-00022.safetensors5.34 GB0f7b724c1ae85e13443e462878208c9924e5baf47cf56670801384ca1e589cc2
model-00017-of-00022.safetensors4.96 GB55f2f67618487b8eacf8f5e36758281d7d91b5458c260cb8b6665004dbf5ec69
model-00018-of-00022.safetensors5.33 GB67ba9ee0b88bb5961a14992748c021c87f8a59bd2efd9c65509e36076a6e3229
model-00019-of-00022.safetensors5.33 GB65c34375b932fb13ecfb27c2588f326450e54604616bde870b38e1c8e96070a0
model-00020-of-00022.safetensors4.97 GB1a56d13e3ef8c80bfd1f571c93cab708a7b2b6aaa1dd958df4fd51f90d0d1eaf
model-00021-of-00022.safetensors5.34 GB42b36361206e0bfbbbf24dba154d28d4fc46163d22aac3edb90574ee1dd8fb58
model-00022-of-00022.safetensors2.08 GB6388ae50b68ffffc8c46e568345ffa95d31d9b647fcc64c2670f15c8549e602a
model.safetensors.index.json422 KB3581f8d40a330d40009d0417359f5f75b7cf79e9f8fc48ba0b1461eabd43dd5f
preprocessor_config.json390 B27225450ac9c6529872ee1924fcb0962ff5634834f817040f444118116f4e516
processor_config.json991 B45fc17c8dd2474af6b493b52483c26c0584b0082d368c480f9fa611e73070040
tokenizer.json20.0 MB06b9509352d2af50381ab2247e083b80d32d5c0aba91c272ca9ff729b6a0e523
tokenizer_config.json1.17 KB792fa3f0cb88b111e54ef3134c873531008c4df471d108da17903426e308aa7b
video_preprocessor_config.json385 B7768af27c1fafa9cc9011c1dc20067e03f8915e03b63504550e11d5066986d13
vocab.json6.72 MBce99b4cb2983d118806ce0a8b777a35b093e2000a503ebde25853284c9dfa003
VersionsEach time the pinned upstream commit or file list changed in this repo. Newest first.
Upstream commitDateContents
07b5dc6c54 pinned now2026-11-14 21:30 UTC34 files · 112 GB

Use it

# custom terms: read the license before you download
$ fw get mlx-community/Qwen3.8-Flash-Next-4bit
# manifest sha256 de1ed434dcb6… · 112 GB · 34 files
$ fw verify ./Qwen3.8-Flash-Next-4bit mlx-community/Qwen3.8-Flash-Next-4bit
Custom license (custom). Read the license text first. Open this panel to show the commands once you accept its terms.
fw.pyChecks the signature and every sha256, tries each source in order, resumes.
$ python3 fw.py get https://freeweights.allternit.com/manifests/mlx-community/Qwen3.8-Flash-Next-4bit.json --dir ./Qwen3.8-Flash-Next-4bit $ python3 fw.py verify ./Qwen3.8-Flash-Next-4bit ./mlx-community__Qwen3.8-Flash-Next-4bit.json
hfThe Hugging Face CLI with one setting. Same repo id.
$ HF_ENDPOINT=https://freeweights.allternit.com hf download mlx-community/Qwen3.8-Flash-Next-4bit --revision 07b5dc6c54600a359b87f1e53e7adf6351c72a2c --local-dir ./Qwen3.8-Flash-Next-4bit
transformersLoad the weights in Python.
import os os.environ["HF_ENDPOINT"] = "https://freeweights.allternit.com" from transformers import AutoModelForImageTextToText, AutoProcessor processor = AutoProcessor.from_pretrained("mlx-community/Qwen3.8-Flash-Next-4bit", revision="07b5dc6c54600a359b87f1e53e7adf6351c72a2c") model = AutoModelForImageTextToText.from_pretrained("mlx-community/Qwen3.8-Flash-Next-4bit", revision="07b5dc6c54600a359b87f1e53e7adf6351c72a2c", device_map="auto")
vLLMServe an OpenAI-compatible API.
$ HF_ENDPOINT=https://freeweights.allternit.com vllm serve mlx-community/Qwen3.8-Flash-Next-4bit --revision 07b5dc6c54600a359b87f1e53e7adf6351c72a2c

License

Name
LicenseRef-Qwen-Community-1.0
Id
LicenseRef-Qwen-Community-1.0
Class
Custom terms
Redistribution
click-through
Listing
Hosted (signed)
Decided
Inherited from the base model Qwen/Qwen3.8-Flash-Next: the strictest license along the lineage applies. 2026-11-14
Text in
README.md
Text sha256
eb54592ce293006d75d181be8dca197282f3b76148852c9c505fc9242bcef3fe
Terms gate
Yes. The commands stay folded until you open them after reading 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: mlx-community <huggingface.co/mlx-community>Newsgroups: alt.freeweights.qwenSubject: Qwen3.8-Flash-Next-4bitDate: 2026-11-14Commit: 07b5dc6Signed: 2DC28CC7B26A283B ✓
Upstream
mlx-community/Qwen3.8-Flash-Next-4bit
Pinned commit
07b5dc6c54600a359b87f1e53e7adf6351c72a2c
Manifest
Qwen3.8-Flash-Next-4bit.json · .minisig
Manifest sha256
de1ed434dcb6ed1637fdd7b309a765b3578e1553448164941e591da5821efe7c
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-Flash-Next-4bit.json $ curl -O https://freeweights.allternit.com/manifests/mlx-community/Qwen3.8-Flash-Next-4bit.json.minisig $ minisign -Vm Qwen3.8-Flash-Next-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

Parent

Qwen/Qwen3.8-Flash-Next · this model is a quantization of it

No model in the catalog names this one as its parent.

Get it
Read license, then download
$ HF_ENDPOINT=https://freeweights.allternit.com hf download mlx-community/Qwen3.8-Flash-Next-4bit
Open in Freeweights Studio

Pull, verify and run it in the Allternit app on Desktop, web or phone.

Signed manifest
minisign, key 2DC28CC7B26A283B. Qwen3.8-Flash-Next-4bit.json. Provenance. Takedown history: none.
Source
huggingface.co/mlx-community/Qwen3.8-Flash-Next-4bit

Related models

All of alt.freeweights.qwen →
From: BAAI <huggingface.co/BAAI>Subject: Aquila-VL-2B-llava-qwen
From: Perciqa <huggingface.co/Perciqa>Subject: Aurora-Code-1
From: BAAI <huggingface.co/BAAI>Subject: bge-reasoner-embed-qwen3-8b-0923
From: amazon <huggingface.co/amazon>Subject: BMOJOF-primed-HQwen3-8B-Instruct
From: nbeerbower <huggingface.co/nbeerbower>Subject: Bubba-Qwen3.6-27B-LoRA
From: bartowski <huggingface.co/bartowski>Subject: bytkim_Qwen3.8-27B-pi-GGUF