DeepSeek-V4-Flash-Spark-Mini GGUF (Q3-Dynamic REAP)

GGUF quantization of deepseek-ai/DeepSeek-V4-Flash-0731 (MIT), Mini variant of the REAP-pruned Spark lineage with dynamic per-layer Q3 allocation.

Files

• DeepSeek-V4-Flash-Spark-Mini-Q3-Dynamic-REAP-ds4.gguf — single-file GGUF
• DeepSeek-V4-Flash-Spark-Mini-Q3-Dynamic-REAP-ds4.gguf.sha256 — checksum sidecar

Usage

Compatible with llama.cpp and other GGUF loaders:

llama-server -m DeepSeek-V4-Flash-Spark-Mini-Q3-Dynamic-REAP-ds4.gguf

License

MIT, inherited from the upstream DeepSeek-V4-Flash-0731 repository. See the upstream model card for full terms.

Acknowledgements

• DeepSeek for the upstream DeepSeek-V4-Flash-0731 (MIT) this quantization derives from.
• The llama.cpp project and community for the GGUF format and runtime support.