Open-weight model
DeepSeek-V3.2-NVFP4-mixed-CT-AutoRound
by INC Optimized Models 2 INCModel2/DeepSeek-V3.2-NVFP4-mixed-CT-AutoRound
DeepSeek-V3.2-NVFP4-mixed-CT-AutoRound is an open-weight model from INC Optimized Models 2. It has 386.3B parameters and a 163,840-token context. At 16-bit it needs about 927 GB of GPU memory, which fits on 4x MI325X from $8.00 an hour; at 4-bit, 231.8 GB on 1x MI325X from $2.00, at the lowest prices in the SAVRN Index. It draws 77 downloads a month.
Parameters386.3B
Context163,840
Weights485.7 GB
License—
AccessOpen weights
Monthly Downloads77
Runs On
What it takes to serve DeepSeek-V3.2-NVFP4-mixed-CT-AutoRound (386.3B parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them.
| Precision | Weights | Memory needed | Cheapest setup | Per hour | Also fits |
|---|---|---|---|---|---|
| 16-bit | 772.5 GB | 927.0 GB | 4x MI325X (256 GB) Vultr |
$8.00 | 5x MI300X $9.25 · 4x MI355X $10.36 |
| 8-bit | 386.3 GB | 463.5 GB | 2x MI325X (256 GB) Vultr |
$4.00 | 2x MI355X $5.18 · 3x MI300X $5.55 |
| 4-bit | 193.1 GB | 231.8 GB | 1x MI325X (256 GB) Vultr |
$2.00 | 1x MI355X $2.59 · 2x MI300X $3.70 |
Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the SAVRN Index, read Sep 18, 2026.
Model Card
The publisher has not written a card for this model.
Configuration
- Architecture
- DeepseekV32ForCausalLM
- Context length (tokens)
- 163,840
- Layers
- 61
- Hidden size
- 7,168
- Feed-forward size
- 18,432
- Attention heads
- 128
- Key/value heads
- 128
- Head dimension
- 64
- Vocabulary size
- 129,280
- Routed experts
- 256
- Experts active per token
- 8
- Model type
- deepseek_v32
- Quantization
- compressed-tensors
Identity and Version
- Repository
- INCModel2/DeepSeek-V3.2-NVFP4-mixed-CT-AutoRound
- Publisher
- INC Optimized Models 2
- Task
- Not stated by the source
- Modality
- Other
- Library
- Not stated by the source
- Parameters
- 386.3B parameters
- Languages
- Not stated by the source
- Revision
- b8e0b8a533224eea74f95682db9281a50786b0de
- First published
- 2026-08-22
- Last updated
- 2026-09-18
Files and Weights
98 files, 485.7 GB in total. The weights are 91 files totalling 485.7 GB in safetensors.
Weights91 files · 485.7 GB
Configuration4 files · 22.8 MB
Tokenizer2 files · 10.0 MB
Repository1 file · 1.6 KB
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| model-00001-of-00072.safetensors | Weights | 5.4 GB | 5710c8264074 |
| model-00001-of-00073.safetensors | Weights | 5.4 GB | 246663c0a608 |
| model-00002-of-00072.safetensors | Weights | 5.4 GB | 84a012cfcdfa |
| model-00002-of-00073.safetensors | Weights | 5.4 GB | a84e4ab62511 |
| model-00003-of-00072.safetensors | Weights | 5.4 GB | 3a14b7613f1b |
| model-00003-of-00073.safetensors | Weights | 5.4 GB | 5f77e062f0a7 |
| model-00004-of-00072.safetensors | Weights | 5.4 GB | 88cd9ca36b22 |
| model-00004-of-00073.safetensors | Weights | 5.4 GB | 703893ce4b57 |
| model-00005-of-00072.safetensors | Weights | 5.4 GB | fa99471a030e |
| model-00005-of-00073.safetensors | Weights | 5.4 GB | d663323f7c6e |
| model-00006-of-00072.safetensors | Weights | 5.4 GB | 0941bc80a7e2 |
| model-00006-of-00073.safetensors | Weights | 5.4 GB | 1ac4e9ac8425 |
| model-00007-of-00072.safetensors | Weights | 5.4 GB | 148227f99156 |
| model-00007-of-00073.safetensors | Weights | 5.4 GB | 5a41cf47b6a9 |
| model-00008-of-00072.safetensors | Weights | 5.4 GB | a747c44d563e |
| model-00008-of-00073.safetensors | Weights | 5.4 GB | c54ebba4c623 |
| model-00009-of-00072.safetensors | Weights | 5.4 GB | b732b4a5325e |
| model-00009-of-00073.safetensors | Weights | 5.3 GB | 177b7ab470f3 |
| model-00010-of-00072.safetensors | Weights | 5.4 GB | acae094bb173 |
| model-00010-of-00073.safetensors | Weights | 5.4 GB | 55b5b431d228 |
| model-00011-of-00072.safetensors | Weights | 5.4 GB | 6bbab467f6bf |
| model-00011-of-00073.safetensors | Weights | 5.4 GB | 96a1ad94f07c |
| model-00012-of-00072.safetensors | Weights | 5.4 GB | 4f2cd17fde23 |
| model-00012-of-00073.safetensors | Weights | 5.4 GB | 8638c35a9aa6 |
| model-00013-of-00072.safetensors | Weights | 5.4 GB | 4b42ef7242ec |
| model-00013-of-00073.safetensors | Weights | 5.4 GB | b95a30c277ab |
| model-00014-of-00072.safetensors | Weights | 5.4 GB | 884389bcf978 |
| model-00014-of-00073.safetensors | Weights | 5.4 GB | 529b76a32064 |
| model-00015-of-00072.safetensors | Weights | 5.4 GB | bbda52bc05bf |
| model-00015-of-00073.safetensors | Weights | 5.4 GB | 112bcfbf164e |
| model-00016-of-00072.safetensors | Weights | 5.4 GB | 5c6ab208639f |
| model-00016-of-00073.safetensors | Weights | 5.4 GB | dce556939ff9 |
| model-00017-of-00072.safetensors | Weights | 5.4 GB | fd66ecdb9f2b |
| model-00017-of-00073.safetensors | Weights | 5.4 GB | dd0e0ec007b7 |
| model-00018-of-00072.safetensors | Weights | 5.4 GB | 6a0e4c7e03ab |
| model-00018-of-00073.safetensors | Weights | 5.4 GB | e23e731dbebf |
| model-00019-of-00073.safetensors | Weights | 5.4 GB | 277b60bf61a9 |
| model-00020-of-00073.safetensors | Weights | 5.4 GB | ba889ce19306 |
| model-00021-of-00073.safetensors | Weights | 5.4 GB | 4cd9e1e2a7f0 |
| model-00022-of-00073.safetensors | Weights | 5.4 GB | a2fa01fe2312 |
| model-00023-of-00073.safetensors | Weights | 5.4 GB | acbd8b0b8708 |
| model-00024-of-00073.safetensors | Weights | 5.4 GB | f48d94b40176 |
| model-00025-of-00073.safetensors | Weights | 5.4 GB | cdee40e6166f |
| model-00026-of-00073.safetensors | Weights | 5.4 GB | 2cc6b8184600 |
| model-00027-of-00073.safetensors | Weights | 5.4 GB | 87ddd74b70a2 |
| model-00028-of-00073.safetensors | Weights | 5.4 GB | 807335dc88c5 |
| model-00029-of-00073.safetensors | Weights | 5.4 GB | a955ddc165cf |
| model-00030-of-00073.safetensors | Weights | 5.4 GB | 2f09dab35fa1 |
| model-00031-of-00073.safetensors | Weights | 5.4 GB | 2f322bd0a244 |
| model-00032-of-00073.safetensors | Weights | 5.4 GB | c560baca178c |
| model-00033-of-00073.safetensors | Weights | 5.4 GB | 8c8059ca5bd3 |
| model-00034-of-00073.safetensors | Weights | 5.4 GB | 0270f59ce935 |
| model-00035-of-00073.safetensors | Weights | 5.4 GB | aff0353b14df |
| model-00036-of-00073.safetensors | Weights | 5.4 GB | 339257107b9b |
| model-00037-of-00073.safetensors | Weights | 5.4 GB | 24ab8d55cf74 |
| model-00038-of-00073.safetensors | Weights | 5.4 GB | 0e55a042a5b0 |
| model-00039-of-00073.safetensors | Weights | 5.4 GB | b0c5877953e6 |
| model-00040-of-00073.safetensors | Weights | 5.4 GB | 04ee20214a3f |
| model-00041-of-00073.safetensors | Weights | 5.4 GB | a4b5ab7acd9a |
| model-00042-of-00073.safetensors | Weights | 5.4 GB | e53aaf44c4d3 |
| model-00043-of-00073.safetensors | Weights | 5.4 GB | 5f1ca0786126 |
| model-00044-of-00073.safetensors | Weights | 5.4 GB | 94aca8521732 |
| model-00045-of-00073.safetensors | Weights | 5.4 GB | 2ad62fa8e0c8 |
| model-00046-of-00073.safetensors | Weights | 5.4 GB | 17b9dd2b0d43 |
| model-00047-of-00073.safetensors | Weights | 5.4 GB | f593e5ae2341 |
| model-00048-of-00073.safetensors | Weights | 5.4 GB | b2693aa1774c |
| model-00049-of-00073.safetensors | Weights | 5.4 GB | e468f9d813e4 |
| model-00050-of-00073.safetensors | Weights | 5.4 GB | 24a5673f83c4 |
| model-00051-of-00073.safetensors | Weights | 5.4 GB | fc450869d17d |
| model-00052-of-00073.safetensors | Weights | 5.4 GB | ab75ce62245a |
| model-00053-of-00073.safetensors | Weights | 5.4 GB | d1ffbe7b76cc |
| model-00054-of-00073.safetensors | Weights | 5.4 GB | d0f157ea3ee2 |
| model-00055-of-00073.safetensors | Weights | 5.4 GB | 26d3745de068 |
| model-00056-of-00073.safetensors | Weights | 5.4 GB | 098dd1db0a19 |
| model-00057-of-00073.safetensors | Weights | 5.4 GB | 286b62021e6e |
| model-00058-of-00073.safetensors | Weights | 5.4 GB | 986aa67046c9 |
| model-00059-of-00073.safetensors | Weights | 5.4 GB | 8d2c25361a72 |
| model-00060-of-00073.safetensors | Weights | 5.4 GB | e5aa84ab883f |
| model-00061-of-00073.safetensors | Weights | 5.4 GB | 8eec7b6a007e |
| model-00062-of-00073.safetensors | Weights | 5.3 GB | d311801bfb4b |
| model-00063-of-00073.safetensors | Weights | 5.4 GB | 8b23289a8d45 |
| model-00064-of-00073.safetensors | Weights | 5.4 GB | 0392bb41b71c |
| model-00065-of-00073.safetensors | Weights | 5.4 GB | 3cc7ce19b770 |
| model-00066-of-00073.safetensors | Weights | 5.4 GB | e074b9c72803 |
| model-00067-of-00073.safetensors | Weights | 5.4 GB | b5664c485138 |
| model-00068-of-00073.safetensors | Weights | 5.4 GB | 15793ccaba37 |
| model-00069-of-00073.safetensors | Weights | 5.4 GB | 3aeed1682053 |
| model-00070-of-00073.safetensors | Weights | 5.4 GB | 0c40ed1ce3c0 |
| model-00071-of-00073.safetensors | Weights | 5.4 GB | af6b90d37e7f |
| model-00072-of-00073.safetensors | Weights | 4.5 GB | d054dfa141de |
| model-00073-of-00073.safetensors | Weights | 3.7 GB | 6807736d43bb |
| config.json | Configuration | 2.4 MB | — |
| generation_config.json | Configuration | 171 B | — |
| model.safetensors.index.json | Configuration | 18.0 MB | 7b2e874910cc |
| quantization_config.json | Configuration | 2.3 MB | — |
| .gitattributes | Repository | 1.6 KB | — |
| tokenizer.json | Tokenizer | 10.0 MB | — |
| tokenizer_config.json | Tokenizer | 396 B | — |
License and Download
- License
- Not stated by the source
- Access
- Open weights, no gate
- Download size
- 485.7 GB
Download from INC Optimized Models 2
Released by INC Optimized Models 2 through its official repository on Hugging Face.
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 485.7 GB |
| 16-bit | 772.5 GB |
| 8-bit | 386.3 GB |
| 4-bit | 193.1 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About DeepSeek-V3.2-NVFP4-mixed-CT-AutoRound
How much GPU memory does DeepSeek-V3.2-NVFP4-mixed-CT-AutoRound need?
About 927 GB at 16-bit and 231.8 GB at 4-bit: the weights (386.3B parameters) plus a working margin. A long context needs more.
What is the cheapest GPU to run DeepSeek-V3.2-NVFP4-mixed-CT-AutoRound on?
At 16-bit, 4x MI325X from $8.00 an hour; at 4-bit, 1x MI325X from $2.00 an hour, at the lowest on-demand prices the SAVRN Index lists.
What is DeepSeek-V3.2-NVFP4-mixed-CT-AutoRound's context length?
163,840 tokens, from the maximum position embeddings in its published configuration.