Daniel Han(@danielhanchen):We worked on a llama.cpp impl for DS4.1-Flash + made Unsloth dynamic GGUFs for it! The engram / PLE can be swapped out from MXFP8 to Q6_K to Q5_K and IQ4_NL precision - we found Q6_K approx = MXFP8 and IQ4_NL loses a tiny bit of accuracy. The engram can stream fully from disk!

Daniel Han

@danielhanchen

Building @UnslothAI • Making open-source LLMs faster, better & more accessible • YC S24 • ex-NVIDIA ML

Joined April 2016

2K Following 37.2K Followers

Daniel Han@danielhanchen

2026.10.19 16:13

We worked on a llama.cpp impl for DS4.1-Flash + made Unsloth dynamic GGUFs for it! The engram / PLE can be swapped out from MXFP8 to Q6_K to Q5_K and IQ4_NL precision - we found Q6_K approx = MXFP8 and IQ4_NL loses a tiny bit of accuracy. The engram can stream fully from disk!

Unsloth AI@UnslothAI

2026.10.19 16:01

DeepSeek V4.1-Flash-1017 can now run locally! 🐳 We made 1-bit retain ~88.2% accuracy while shrinking size from 508GB to 250GB (−52%). Run on a 128GB+ RAM device. V4.1 Flash 1017 is the strongest open model to date Guide: GGUF: