Open-weight model
DeepSeek-V4.1-Flash-7layers-compact-engram
by Xiaochang Wu xwu-intel/DeepSeek-V4.1-Flash-7layers-compact-engram
DeepSeek-V4.1-Flash-7layers-compact-engram is an open-weight model from Xiaochang Wu, released under Apache License 2.0. It has 112.5B parameters and a 1,048,576-token context. At 16-bit it needs about 270.1 GB of GPU memory, which fits on 1x MI355X from $2.59 an hour; at 4-bit, 67.5 GB on 1x MI300X from $1.85, at the lowest prices in the SAVRN Index.
Runs On
What it takes to serve DeepSeek-V4.1-Flash-7layers-compact-engram (112.5B 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 | 225.0 GB | 270.1 GB | 1x MI355X (288 GB) Vultr |
$2.59 | 2x MI300X $3.70 · 2x MI325X $4.00 |
| 8-bit | 112.5 GB | 135.0 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x MI325X $2.00 · 1x MI355X $2.59 |
| 4-bit | 56.3 GB | 67.5 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
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 Oct 18, 2026.
Model Card
The publisher has not written a card for this model.
Configuration
- Architecture
- DeepseekV41ForCausalLM
- Context length (tokens)
- 1,048,576
- Layers
- 7
- Hidden size
- 5,120
- Attention heads
- 64
- Key/value heads
- 1
- Head dimension
- 512
- Vocabulary size
- 129,280
- Routed experts
- 384
- Experts active per token
- 6
- Sliding window (tokens)
- 128
- RoPE base
- 10,000
- Model type
- deepseek_v41
- Quantization
- fp8
Identity and Version
- Repository
- xwu-intel/DeepSeek-V4.1-Flash-7layers-compact-engram
- Publisher
- Xiaochang Wu
- Task
- Not stated by the source
- Modality
- Other
- Library
- Not stated by the source
- Parameters
- 112.5B parameters
- Languages
- Not stated by the source
- Revision
- bd54c94438df43deb78c2c87891310f750c4fd55
- First published
- 2026-09-29
- Last updated
- 2026-09-29
Files and Weights
20 files, 63.5 GB in total. The weights are 14 files totalling 63.5 GB in safetensors.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| model-00001-of-00014.safetensors | Weights | 970.5 MB | 886aebdafa08 |
| model-00002-of-00014.safetensors | Weights | 1.3 GB | 4320066fc695 |
| model-00003-of-00014.safetensors | Weights | 7.4 GB | e285bb9a20ce |
| model-00004-of-00014.safetensors | Weights | 7.4 GB | 12e4736ae5b3 |
| model-00005-of-00014.safetensors | Weights | 7.4 GB | 46c32a800b9c |
| model-00006-of-00014.safetensors | Weights | 7.4 GB | db94d207ab7a |
| model-00007-of-00014.safetensors | Weights | 7.4 GB | 581202c131f3 |
| model-00008-of-00014.safetensors | Weights | 7.4 GB | 8809e9ecd292 |
| model-00009-of-00014.safetensors | Weights | 7.4 GB | 1ac2b61b2e35 |
| model-00010-of-00014.safetensors | Weights | 1.3 GB | d762b688f138 |
| model-00011-of-00014.safetensors | Weights | 2.7 GB | 9a6b39fb88a2 |
| model-00012-of-00014.safetensors | Weights | 2.6 GB | 0cc9d5f6ca3a |
| model-00013-of-00014.safetensors | Weights | 2.7 GB | e625902027b9 |
| model-00014-of-00014.safetensors | Weights | 184.1 MB | 5e5e4442c03a |
| config.json | Configuration | 2.9 KB | — |
| model.safetensors.index.json | Configuration | 1.5 MB | — |
| README.md | Documentation | 28 B | — |
| .gitattributes | Repository | 1.5 KB | — |
| tokenizer.json | Tokenizer | 6.4 MB | — |
| tokenizer_config.json | Tokenizer | 801 B | — |
License and Download
- License
- apache-2.0
- Access
- Open weights, no gate
- Download size
- 63.5 GB
Released by Xiaochang Wu through its official repository on Hugging Face. Read the license.
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 63.5 GB |
| 16-bit | 225.0 GB |
| 8-bit | 112.5 GB |
| 4-bit | 56.3 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About DeepSeek-V4.1-Flash-7layers-compact-engram
How much GPU memory does DeepSeek-V4.1-Flash-7layers-compact-engram need?
About 270.1 GB at 16-bit and 67.5 GB at 4-bit: the weights (112.5B parameters) plus a working margin. A long context needs more.
What is the cheapest GPU to run DeepSeek-V4.1-Flash-7layers-compact-engram on?
At 16-bit, 1x MI355X from $2.59 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.
Can I use DeepSeek-V4.1-Flash-7layers-compact-engram commercially?
Yes. DeepSeek-V4.1-Flash-7layers-compact-engram is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.
What is DeepSeek-V4.1-Flash-7layers-compact-engram's context length?
1,048,576 tokens, from the maximum position embeddings in its published configuration.