DeepSeek V4 Flash 180B ROCmFP4 STRIX LEAN needs about 97.7 GB of memory at MXFP4 (a 96.4 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 2 quants, from 96.4 GB (MXFP4) to 98.7 GB (MXFP4). Longer context needs more: at 32K tokens the KV cache alone adds about 2.9 GB.
DeepSeek V4 Flash 180B ROCmFP4 STRIX LEAN – 180B parameter LLM
DeepSeek V4 Flash 180B ROCmFP4 STRIX LEAN is a 180B parameter model with speculative decoding support. It is quantized to MXFP4 and available in two sharded variants. The model is licensed under Apache‑2.0 and supports reasoning and tool use.
Good at
- Reasoning
- Tool calling
- Speculative decoding
Limits
- Requires ROCmFP4 build of llama.cpp
- No vision support
- Large VRAM needed
Best for
- Agentic applications
- Complex reasoning tasks
- Tool‑enabled workflows
Who it’s for. Developers building agentic systems on AMD unified‑memory GPUs. The model is suitable for users who need large context and speculative decoding capabilities.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| MXFP4 | 96.4 GB (3 parts) | 97.7 GB | Too big | Get file → | |
| MXFP4 | 98.7 GB (3 parts) | 99.9 GB | Too big | Get file → |
For 8 GB VRAM, use the 96.43 GB MXFP4 variant; it estimates 97.7 GB memory at 8 k context. For 16 GB VRAM, the 98.68 GB variant uses about 99.9 GB memory. Both are sharded into 3 parts to stay under the 50 GB per‑file limit.
What GPU do you need for DeepSeek V4 Flash 180B ROCmFP4 STRIX LEAN?
These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.
- 4 GB VRAM: no quant fits at an 8K context.
- 6 GB VRAM: no quant fits at an 8K context.
- 8 GB VRAM: no quant fits at an 8K context.
- 12 GB VRAM: no quant fits at an 8K context.
- 16 GB VRAM: no quant fits at an 8K context.
- 24 GB VRAM: no quant fits at an 8K context.
- 32 GB VRAM: no quant fits at an 8K context.
- 48 GB VRAM: no quant fits at an 8K context.
Which DeepSeek V4 Flash 180B ROCmFP4 STRIX LEAN quant should you choose?
MXFP4 (96.4 GB) is the middle option in this repository. Lower bit counts mean smaller files and more quality loss.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about DeepSeek V4 Flash 180B ROCmFP4 STRIX LEAN
How much VRAM does DeepSeek V4 Flash 180B ROCmFP4 STRIX LEAN need?
About 97.7 GB at MXFP4 with an 8K context window. That is the 96.4 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.
Can I run DeepSeek V4 Flash 180B ROCmFP4 STRIX LEAN on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 97.7 GB, so you would need a larger GPU or CPU offloading.
What is the file size of DeepSeek V4 Flash 180B ROCmFP4 STRIX LEAN GGUF?
It depends on the quant: MXFP4 is 96.4 GB, MXFP4 is 98.7 GB.
What context length does DeepSeek V4 Flash 180B ROCmFP4 STRIX LEAN support?
The GGUF header lists a trained context length of 1,048,576 tokens (1024K). The KV cache grows with the context you use: about 0.7 GB at 8K and 2.9 GB at 32K.
Where can I download DeepSeek V4 Flash 180B ROCmFP4 STRIX LEAN GGUF?
The files are on Hugging Face in the kingjones777/DeepSeek-V4-Flash-180B-ROCmFP4-STRIX_LEAN-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does DeepSeek V4 Flash 180B ROCmFP4 STRIX LEAN use?
Hugging Face lists the license as apache-2.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.
Can I run this model on 8 GB VRAM?
No. The smallest variant requires about 97.7 GB VRAM at 8 k context, which exceeds 8 GB.
Does this model support vision?
No. The model has no vision capabilities.
What build of llama.cpp is needed?
A ROCmFP4‑enabled build compiled for gfx1151 is required to load the Q4_0_ROCMFP4 quantizations.