alt.freeweights.deepseek / DeepSeek-V4.1-Flash-1017-oQ2e-ngram-mtp

From:

mlx-community <huggingface.co/mlx-community>Newsgroups: alt.freeweights.deepseekSubject:

DeepSeek-V4.1-Flash-1017-oQ2e-ngram-mtp

Date: 2026-11-15Commit: a5cebedSigned: 2DC28CC7B26A283B ✓

Image text to text model with 254.8B parameters. Safetensors weights, 393 GB in all. Published by mlx-community under MIT.

254.8B params

393GBDeepSeek-V4.1-Flash-1017-oQ2e-ngram-mtpsafetensorspermissive

FactsParams254.8BSize393 GBFormatssafetensorsTaskimage text to textLicenseMITFrom sourcepermissive

$ fw get mlx-community/DeepSeek-V4.1-Flash-1017-oQ2e-ngram-mtp
# manifest sha256 06683100a98d… · 393 GB · 53 files
$ fw verify ./DeepSeek-V4.1-Flash-1017-oQ2e-ngram-mtp mlx-community/DeepSeek-V4.1-Flash-1017-oQ2e-ngram-mtp

From sourceHosted (signed)mlx-community/DeepSeek-V4.1-Flash-1017-oQ2e-ngram-mtpLast check: 2026-11-15 05:07 UTC

Get help

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

Quantized of DeepSeek-V4.1-Flash-1017.

Works with

Run23 Fine-tune29 Convert15 Evaluate13 Scan13 Apps62

Runs with

Ranked by fit to this model (safetensors), then by GitHub stars.

vLLM exo Text Generation WebUI

SGLang MLC LLM KTransformers LMCache Xinference LMDeploy mistral.rs Mooncake FlashInfer

Show all 23

AIBrix fastllm ExLlamaV3

Transformers LocalAI tinygrad Triton Inference Server BentoML CTranslate2 ExLlamaV2 LitServe

Convert with

mergekit TensorRT Model Optimizer

LLM Compressor Neural Compressor AutoRound AQLM GPTQModel ONNX Core ML Tools ExecuTorch

Show all 15

Evaluate with

Opik DeepEval Ragas lm-evaluation-harness OpenCompass SWE-bench

simple-evals TruLens HELM

LightEval

Evaluate Weave

Show all 13

Autoevals

Scan with

promptfoo garak Guardrails NeMo Guardrails Giskard DeepTeam AI Fairness 360 Fairlearn Detoxify MCP Scanner AgentDojo ModelScan

Show all 13

picklescan

Overview Model card Files Use it License Provenance Model tree

Overview

Size: 393 GB
Weights: 393 GB (safetensors)
Parameters: 254.8B
Formats: safetensors
Task: image text to text
License: MIT (permissive)
Files: 53, each with a sha256
Publisher: mlx-community
Added: 2026-11-15
Section: alt.freeweights.deepseek

Lineage

Original release from the publisher.

Base model: deepseek-ai/DeepSeek-V4.1-Flash-1017

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
From source	huggingface.co	2026-11-15 05:07 UTC	passed

Model card

From the publisher at commit a5cebed. Shown as plain text with links; images open at the publisher.

DeepSeek-V4.1-Flash-1017-oQ2e-ngram-mtp#

N-gram (Engram) tables are kept at the original FP8 precision. All other weights match DeepSeek-V4.1-Flash-1017-oQ2e-mtp, so resident memory is the same with Engram SSD offload.

This model was quantized using oQ (oMLX v0.7.1) mixed-precision quantization with Improved oQe Quantization (jundot/omlx#4385).

A mixed-precision MLX checkpoint of deepseek-ai/DeepSeek-V4.1-Flash-1017, a minor update to DeepSeek-V4.1-Flash with improved reasoning, coding, agentic and visual performance. It is a multimodal mixture-of-experts model with 552B backbone parameters, a Causal Encoder-Decoder layout, Compressed Sparse Attention 2 (CSA2), Engram conditional memory and DSpark speculative decoding. It includes the language model, the Engram tables, the vision encoder, the multimodal aligner and the DSpark MTP weights.

The effective average is 2.688 bits/weight excluding the Engram tables, or 4.121 bits/weight for the complete checkpoint.

Quantization and Bit Distribution#

oQ2e allocates bits by measured layer sensitivity, then rounds each group with an importance matrix (1024 samples x 512 tokens) and a weighted least-squares refit of its scale and bias. The Engram tables keep the original FP8 storage, E4M3 values with one E8M0 scale per 32 channels, and are left out of the bit budget because oMLX can serve them from SSD instead of memory.

Storage format	Logical weights	Share of total	Tensor storage	Effective bits/weight
Affine 2-bit, group 64	543.582B	71.224%	158.203 GiB	2.50
MXFP4, group 32	13.590B	1.781%	6.724 GiB	4.25
Affine 8-bit, group 64	5.200B	0.681%	5.145 GiB	8.50
MXFP8, group 32	0.779B	0.102%	0.748 GiB	8.25
BF16	3.399B	0.445%	6.332 GiB	16.00
FP32	0.042B	0.006%	0.158 GiB	32.00
Engram MXFP8, group 32	196.614B	25.762%	188.833 GiB	8.25
Total	763.205B	100%	366.143 GiB	4.121

Storage Breakdown#

Component	Logical weights	Storage
Language backbone	551.881B	181.366 GB / 168.911 GiB
Engram tables	196.614B	202.758 GB / 188.833 GiB
Vision encoder	0.485B	0.971 GB / 0.904 GiB
MTP head	14.225B	8.047 GB / 7.495 GiB
Total	763.205B	393.143 GB / 366.143 GiB

License#

MIT, following the source model. See LICENSE.

Files

53 files · 393 GBWeights 393 GB (safetensors). Sizes and sha256 from the signed manifest.

Show all 53 files with their sha256

File	Size	sha256
.gitattributes	1.52 KB	11ad7efa24975ee4b0c3c3a38ed18737f0658a5f75a0a96787b576a78a023361
LICENSE	1.08 KB	f2c6c602815669d292889e5be8c802f2ed950653b77999b1584e8e6aed25d040
README.md	2.88 KB	c1447c9266e54325a97e46fc74869cb9943917268c404eed9b4668d76af8b9a3
config.json	134 KB	aad9a3f7518ea88493e1ef566ba25b8a7b64a2b8f5e69e0e1336c0582d1681d8
model-00001-of-00045.safetensors	101 GB	640fa5600d2aa35d1f3386e44b2db02be9bab7fba4a11bd6280641f82996f064
model-00002-of-00045.safetensors	101 GB	68d201c011a97da8b2cb9f9098568eb66b1f727473a2bea686ae0591f46f7aca
model-00003-of-00045.safetensors	4.01 GB	e65f647085e2b5ec750efca6968e38a3c91ab3a1e372b258394fc60b4de999d5
model-00004-of-00045.safetensors	4.63 GB	2cf368e1d9ede0ba1a81fa97b58578836edca6eb765d18246553a755dd17e6ae
model-00005-of-00045.safetensors	4.46 GB	99c36f9bc56b82821801647028bb53bd5af0e866fcc39ad7abef362ccfe02da6
model-00006-of-00045.safetensors	4.46 GB	122087957620c1d92f78471930e13c6717f698af6b794fd38f7533f291997473
model-00007-of-00045.safetensors	4.46 GB	49a8f7bbe571285df6609a93e0fc9323752aff5670d89bc0f8c096c1f3af89f0
model-00008-of-00045.safetensors	4.37 GB	a8f5f18c5c59cdc3e3a43d08a317fd95b5db8e1fe58ac6de749ab3d6cdfe1da7
model-00009-of-00045.safetensors	4.46 GB	ab351b2b1686edb5fd4f5cd72288cec693c043ccc21c0e03a78071128263bf96
model-00010-of-00045.safetensors	4.46 GB	d2571d2cacdc977f8cadba108679abb30ebd0d32a282042341f4a33ca3659844
model-00011-of-00045.safetensors	4.46 GB	7ff19d4069ff4d016d4f0f47521bdd2b95806f481c2db77f7f9cbffc95311e22
model-00012-of-00045.safetensors	4.46 GB	ac7c2c361d1301fb5e867dd36b78b318cc9ce0181c7d5a0e2f0ec6dd9416b54b
model-00013-of-00045.safetensors	4.46 GB	c178d80c6fc69ea719de8321e37c9502e9e4ea9f514ee44e483c7a17a05493b1
model-00014-of-00045.safetensors	4.47 GB	c6068523986805d9fba9fa8ff2216d7d42f68972ee4a75075414a46cb3e04ddb
model-00015-of-00045.safetensors	4.46 GB	d44c1a4fd06db4a3bc87074e7c429771a5cf4d2f05e46ea9fa845c56720f1f63
model-00016-of-00045.safetensors	4.46 GB	3fedc7d1279b2593513cf0bb55d005885c62815fee5fd9aaac1b18050cc50831
model-00017-of-00045.safetensors	4.46 GB	0aa45d0d8f9cf34d751f1f210ded602a4c4bd20dd2ea33bb0834a57d901ec0a6
model-00018-of-00045.safetensors	4.53 GB	74770d257ef8eb4225f4fdef7eaab4f474952976af05fab52879ff5cda78c662
model-00019-of-00045.safetensors	4.46 GB	c3022962c1f99b25ea1e522868f8a0c4c8d8e2c5bebfe97b92216218ebc92825
model-00020-of-00045.safetensors	4.46 GB	18f3763258913db244e9b99efe4ee8102f01dd55bdbd16ca2b3675af3232f449
model-00021-of-00045.safetensors	4.46 GB	4d68dc3e228c7918af4230e7261c67d4a4a462ff2992d82ae83f833d0a9d6572
model-00022-of-00045.safetensors	4.46 GB	ef69718f395594a713da1af3294d76324cf4f31eba3e7cfdd0ea90192e552b18
model-00023-of-00045.safetensors	4.45 GB	8efb1958f6f14e72160e26208952dd905a9636d1b0e2726a21d1e9e59fdef4f4
model-00024-of-00045.safetensors	4.12 GB	04d9f1a858af534f04772f83cc3594f0e119c215d05db48ce4c4adc65a6c7ce9
model-00025-of-00045.safetensors	4.46 GB	1cbce3ab3d95bb44a37f1d47e3c0a9bebbb2c050fb34f9fe8117e1f94f739fc2
model-00026-of-00045.safetensors	4.46 GB	0cb5a2e92ea9c9a338defc53f6dbe360ba17e125aec2faf52a81fb9918b6194d
model-00027-of-00045.safetensors	4.46 GB	bd20798ea45519c3e48115d906b581626cb362ba2b8137584796acf80fcb2214
model-00028-of-00045.safetensors	4.46 GB	43eeec279ea47f546be9a9e78768f5308be7045a40dc1e361dae9245e11fa88f
model-00029-of-00045.safetensors	4.64 GB	5531cef93f9b4b56c1a11cd38f33aa5ab7f991cb06f0df8c21304a134bd20615
model-00030-of-00045.safetensors	3.84 GB	a2b096c2f0e167aeca6b9c58548e73b8e92f311a9b849e098bb7c40439419fe6
model-00031-of-00045.safetensors	4.46 GB	73701e35d371e65381949f212a9d1a620886fdb394bb7600168556c12ce351fb
model-00032-of-00045.safetensors	4.46 GB	78e59529be6245d87c51f32edb4e232b7b385b5462b4e53e40e942d294780c96
model-00033-of-00045.safetensors	4.46 GB	19fc0bd2a4a77192e1b1b8e3c5e31044117f3f29d5592967e359906c6229090b
model-00034-of-00045.safetensors	4.46 GB	cae01910793eb9c38b7bae9606fcb59c567b23f123d96eafd7d30744c3ca7cb8
model-00035-of-00045.safetensors	4.37 GB	b37c15df53c494caca4e8641508434f5e5eecaff0c4c7b9f2138e13e50463c40
model-00036-of-00045.safetensors	4.46 GB	9be4fd8ba8e6ff4b704a95733fc4113be35644111706db913e21069e0be5dd18
model-00037-of-00045.safetensors	4.47 GB	2bf64209fa6073ea75cce393e2991795f6f65646450edee2396506ff72f5bb28
model-00038-of-00045.safetensors	4.46 GB	c3e040556777a1b14c66fcf605fcc5caafd4d73cc59ae82e8a5855294be2be93
model-00039-of-00045.safetensors	4.46 GB	19fa5434d53ce37555182824e452492709312a3305bc464e01e495da70930b0f
model-00040-of-00045.safetensors	4.46 GB	5b1b734d6529b4ed0543feffd101ccbb3b95ff26911b8c583d6222781735b71e
model-00041-of-00045.safetensors	4.46 GB	934e8b80745daaf1c080dffa575317b84be0fbdc97d796c3a9a7c6cfe9f9cff3
model-00042-of-00045.safetensors	4.46 GB	8976cb8921812e4f5f51aa69d956f7b8cae53c55cb75d429b5eaf9d62149a3ef
model-00043-of-00045.safetensors	4.46 GB	8b42b1e4b7d04a87f279f28c9dd9fb1a0eacacc868c269fca8e69c021be3ba32
model-00044-of-00045.safetensors	4.46 GB	4eec43bff9de300cdbd6a488fcb60a9a8326dff1778ecdbe702446cb67a739ce
model-00045-of-00045.safetensors	4.25 GB	67a3a326dfb7be2fec0854eacf68293bace11caf74faf53fa1c86b5465e1e75a
model.safetensors.index.json	196 KB	253791cbba66f7487343b7ca40cc9353ef80756105b61ba2702a097dce707270
quantization_report.json	53.4 KB	a5613a444c5fd5bfa4ded3d5eb501f587e556de3ac09f383e8b8ac64ba9d5328
tokenizer.json	6.37 MB	c90dfa01249db1be4245780a052ede752e1361c612ac6d08e2bdada7d599476b
tokenizer_config.json	801 B	6ac8c8dc065ed118161d02dd532749ae3f52c243deac27872134fae2f50d8547

VersionsEach time the pinned upstream commit or file list changed in this repo. Newest first.

Upstream commit	Date	Contents
a5cebed45b pinned now	2026-11-15 05:07 UTC	53 files · 393 GB

Use it

$ fw get mlx-community/DeepSeek-V4.1-Flash-1017-oQ2e-ngram-mtp
# manifest sha256 06683100a98d… · 393 GB · 53 files
$ fw verify ./DeepSeek-V4.1-Flash-1017-oQ2e-ngram-mtp mlx-community/DeepSeek-V4.1-Flash-1017-oQ2e-ngram-mtp

fw.pyChecks the signature and every sha256, tries each source in order, resumes.

$ python3 fw.py get https://freeweights.allternit.com/manifests/mlx-community/DeepSeek-V4.1-Flash-1017-oQ2e-ngram-mtp.json --dir ./DeepSeek-V4.1-Flash-1017-oQ2e-ngram-mtp
$ python3 fw.py verify ./DeepSeek-V4.1-Flash-1017-oQ2e-ngram-mtp ./mlx-community__DeepSeek-V4.1-Flash-1017-oQ2e-ngram-mtp.json

hfThe Hugging Face CLI with one setting. Same repo id.

$ HF_ENDPOINT=https://freeweights.allternit.com hf download mlx-community/DeepSeek-V4.1-Flash-1017-oQ2e-ngram-mtp --revision a5cebed45b736e0de68cea54b1a0922a7426d835 --local-dir ./DeepSeek-V4.1-Flash-1017-oQ2e-ngram-mtp

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/DeepSeek-V4.1-Flash-1017-oQ2e-ngram-mtp", revision="a5cebed45b736e0de68cea54b1a0922a7426d835")
model = AutoModelForImageTextToText.from_pretrained("mlx-community/DeepSeek-V4.1-Flash-1017-oQ2e-ngram-mtp", revision="a5cebed45b736e0de68cea54b1a0922a7426d835", device_map="auto")

vLLMServe an OpenAI-compatible API.

$ HF_ENDPOINT=https://freeweights.allternit.com vllm serve mlx-community/DeepSeek-V4.1-Flash-1017-oQ2e-ngram-mtp --revision a5cebed45b736e0de68cea54b1a0922a7426d835

License

Name: MIT License
Id: MIT
Class: Permissive
Redistribution: yes
Listing: Hosted (signed)
Decided: License as declared by the publisher. No person read it. 2026-11-15
Text in: LICENSE
Text sha256: f2c6c602815669d292889e5be8c802f2ed950653b77999b1584e8e6aed25d040
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:

mlx-community <huggingface.co/mlx-community>Newsgroups: alt.freeweights.deepseekSubject: DeepSeek-V4.1-Flash-1017-oQ2e-ngram-mtpDate: 2026-11-15Commit: a5cebedSigned: 2DC28CC7B26A283B ✓

Upstream: mlx-community/DeepSeek-V4.1-Flash-1017-oQ2e-ngram-mtp
Pinned commit: a5cebed45b736e0de68cea54b1a0922a7426d835
Manifest: DeepSeek-V4.1-Flash-1017-oQ2e-ngram-mtp.json · .minisig
Manifest sha256: 06683100a98d3a43d185eeeff78f1fbcb0990068a75c8713ca976bc31fbf3d58
Signature: minisign, Ed25519, key 2DC28CC7B26A283B
Created: 2026-11-15 05:07 UTC
Generator: freeweights-manifest/0.1.0
Takedowns: None.

How to verify

$ curl -O https://freeweights.allternit.com/manifests/mlx-community/DeepSeek-V4.1-Flash-1017-oQ2e-ngram-mtp.json
$ curl -O https://freeweights.allternit.com/manifests/mlx-community/DeepSeek-V4.1-Flash-1017-oQ2e-ngram-mtp.json.minisig
$ minisign -Vm DeepSeek-V4.1-Flash-1017-oQ2e-ngram-mtp.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

deepseek-ai/DeepSeek-V4.1-Flash-1017 · this model is a quantization of it

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

Get it

Download

$ HF_ENDPOINT=https://freeweights.allternit.com hf download mlx-community/DeepSeek-V4.1-Flash-1017-oQ2e-ngram-mtp

Open in Freeweights Studio

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

Signed manifest

minisign, key 2DC28CC7B26A283B. DeepSeek-V4.1-Flash-1017-oQ2e-ngram-mtp.json. Provenance. Takedown history: none.

Source

huggingface.co/mlx-community/DeepSeek-V4.1-Flash-1017-oQ2e-ngram-mtp

Related models

All of alt.freeweights.deepseek →

From: timm <huggingface.co/timm>Subject: deepseek_vit_412m.deepseek_v4_1_flash

From: bartowski <huggingface.co/bartowski>Subject: deepseek-ai_DeepSeek-R1-0528-GGUF

From: bartowski <huggingface.co/bartowski>Subject: deepseek-ai_DeepSeek-R1-0528-Qwen3-8B-GGUF

From: bartowski <huggingface.co/bartowski>Subject: deepseek-ai_DeepSeek-V3-0324-GGUF

From: bartowski <huggingface.co/bartowski>Subject: deepseek-ai_DeepSeek-V3.1-GGUF

From:

mlx-community <huggingface.co/mlx-community>Subject: deepseek-ai-DeepSeek-V4-Flash-8bit

DeepSeek-V4.1-Flash-1017-oQ2e-ngram-mtp

Works with

Runs with

Fine-tune with

Convert with

Evaluate with

Scan with

Works with

Overview

Model card

DeepSeek-V4.1-Flash-1017-oQ2e-ngram-mtp#

Quantization and Bit Distribution#

Storage Breakdown#

License#

Files

Use it

License

Provenance

How to verify

Model tree

Related models