DeepSeek V4.1 Flash EXL3 3.0bpw RANMA EXL3 needs about 8.6 GB of memory at its main file (a 7.3 GB file) with an 8K context window, so it fits on a 12 GB GPU. This repository has 2 quants, from 7.3 GB to 211.1 GB.
DeepSeek V4.1 Flash EXL3 3.0bpw RANMA EXL3 – 552B multimodal MoE quantized for ranma.cpp
DeepSeek V4.1 Flash EXL3 3.0bpw RANMA EXL3 is a 552B multimodal Mixture‑of‑Experts model quantized with EXL3. It requires ranma.cpp 20261013 or newer and engram tables from the base model. The GGUF files are split into shards and use a 3.02 bpw decoder bitrate.
Good at
- Large context 1,048,576 tokens
- EXL3 quantization for low memory traffic
- MoE architecture with 8B/16B activated per token
Limits
- Requires ranma.cpp 20261013+
- No vision encoder included
- Large memory footprint for full shards
Best for
- High‑capacity inference on RDNA4 GPUs
- Research on MoE quantization
- Deployments needing 1M token context
Who it’s for. Engineers and researchers building inference pipelines for large multimodal models on RDNA4 hardware. It is suitable for those who can supply the 202.8 GB engram tables and run ranma.cpp.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| — | 7.3 GB | 8.6 GB | Too big | Get file → | |
| — | 211.1 GB (6 parts) | 237.0 GB | Too big | Get file → |
For systems with 8 GB VRAM, the 7.25 GB shard requires ~8.6 GB memory at 8 k context, so it is marginal. The 211.13 GB shard needs ~237 GB memory, suitable only for large‑scale setups. Use the 7.25 GB shard if you can allocate slightly more than 8 GB VRAM and host RAM.
What GPU do you need for DeepSeek V4.1 Flash EXL3 3.0bpw RANMA EXL3?
These figures add the file size, an fp16 KV cache at an 8K context window 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: the best quant that fits is the main file (7.3 GB file, about 8.6 GB in use).
- 16 GB VRAM: the best quant that fits is the main file (7.3 GB file, about 8.6 GB in use).
- 24 GB VRAM: the best quant that fits is the main file (7.3 GB file, about 8.6 GB in use).
- 32 GB VRAM: the best quant that fits is the main file (7.3 GB file, about 8.6 GB in use).
- 48 GB VRAM: the best quant that fits is the main file (7.3 GB file, about 8.6 GB in use).
Specifications
Architecture from the GGUF header
Frequently asked questions about DeepSeek V4.1 Flash EXL3 3.0bpw RANMA EXL3
How much VRAM does DeepSeek V4.1 Flash EXL3 3.0bpw RANMA EXL3 need?
About 8.6 GB at its main file with an 8K context window. That is the 7.3 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 V4.1 Flash EXL3 3.0bpw RANMA EXL3 on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 8.6 GB, so you would need a larger GPU or CPU offloading.
What is the file size of DeepSeek V4.1 Flash EXL3 3.0bpw RANMA EXL3 GGUF?
It depends on the quant: file is 7.3 GB, file is 211.1 GB.
What context length does DeepSeek V4.1 Flash EXL3 3.0bpw RANMA EXL3 support?
The GGUF header lists a trained context length of 1,048,576 tokens (1024K).
Where can I download DeepSeek V4.1 Flash EXL3 3.0bpw RANMA EXL3 GGUF?
The files are on Hugging Face in the SaturnHeaven/DeepSeek-V4.1-Flash-EXL3-3.0bpw-RANMA-EXL3-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 V4.1 Flash EXL3 3.0bpw RANMA EXL3 use?
Hugging Face lists the license as mit. 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?
The 7.25 GB shard requires ~8.6 GB memory at 8 k context, so 8 GB VRAM is insufficient.
Does this model support vision input?
No, the vision encoder is not included in this GGUF package.
Where do I get the engram tables?
Download shards 47 and 48 (~95 GiB each) from the base model repository.