---
license: mit
library_name: gguf
tags:
- deepseek
- deepseek-v4
- deepseek-v4-flash
- moe
- reap
- gguf
- mxfp4
---

# DeepSeek-V4-Flash-0731-REAP-K160-GGUF

GGUF quantizations of [0xSero/DeepSeek-V4-Flash-0731-REAP](https://huggingface.co/0xSero/DeepSeek-V4-Flash-0731-REAP) — a REAP expert-pruned DeepSeek-V4-Flash checkpoint retaining 160 of 256 routed experts per MoE layer.

- **Base model:** [0xSero/DeepSeek-V4-Flash-0731-REAP](https://huggingface.co/0xSero/DeepSeek-V4-Flash-0731-REAP)
- **Original model:** [deepseek-ai/DeepSeek-V4-Flash-0731](https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-0731)
- **Format:** GGUF
- **Total params:** ~193B
- **Routed experts:** 160 per layer (REAP-pruned from 256)
- **Architecture:** deepseek4

## Variants

| File | Quant | Size | Description |
|------|-------|------|-------------|
| DeepSeek-V4-Flash-0731-REAP-K160-MXFP4.gguf | MXFP4 (native) | 95 GB | Direct conversion, experts at native 4-bit precision |
| DeepSeek-V4-Flash-0731-REAP-K160-Q4_K_M.gguf | Q4_K_M | 102 GB | K-quant mixing (Q4_K / Q6_K) |
| DeepSeek-V4-Flash-0731-REAP-K160-Q2_K.gguf | Q2_K | ~65 GB | Compact variant for low-memory hardware |

## Serving

These GGUFs use the `deepseek4` architecture. Compatible runtimes:
- DwarfStar (ds4) — reference engine, supports MXFP4 native format
- llama.cpp — with GGML_CUDA=ON for GPU support

## License

MIT — inherited from the base model.

## Acknowledgements

- DeepSeek for the base model
- 0xSero for the REAP pruning
- Cerebras Research for REAP
- llama.cpp contributors
- antirez for DwarfStar