hf_mirrors/DeepSeek-V4.1-Flash-1017-oQ4e-mtp
Model Introduction
Files & versions
Analysis

DeepSeek-V4.1-Flash-1017-oQ4e-mtp

This model was quantized using oQ (oMLX v0.7.1) mixed-precision quantization with Improved oQe Quantization (jundot/omlx#4385).

A mixed-precision MLX checkpoint of deepseek-ai/DeepSeek-V4.1-Flash-1017, a minor update to DeepSeek-V4.1-Flash with improved reasoning, coding, agentic and visual performance. It is a multimodal mixture-of-experts model with 552B backbone parameters, a Causal Encoder-Decoder layout, Compressed Sparse Attention 2 (CSA2), Engram conditional memory and DSpark speculative decoding. It includes the language model, the Engram tables, the vision encoder, the multimodal aligner and the DSpark MTP weights.

The effective average is 4.365 bits/weight excluding the Engram tables, or 4.529 bits/weight for the complete checkpoint.

Quantization and Bit Distribution

oQ4e keeps the original FP4 and FP8 projection precision, which already fits the budget, and quantizes only the Engram tables.

Storage formatLogical weightsShare of totalTensor storageEffective bits/weight
MXFP4, group 32557.171B73.004%275.669 GiB4.25
MXFP8, group 325.959B0.781%5.723 GiB8.25
BF163.419B0.448%6.369 GiB16.00
FP320.042B0.006%0.158 GiB32.00
Engram affine 4-bit, group 32196.614B25.762%114.444 GiB5.00
Total763.205B100%402.363 GiB4.529

Storage Breakdown

ComponentLogical weightsStorage
Language backbone551.881B300.132 GB / 279.520 GiB
Engram tables196.614B122.884 GB / 114.444 GiB
Vision encoder0.485B0.971 GB / 0.904 GiB
MTP head14.225B8.047 GB / 7.495 GiB
Total763.205B432.033 GB / 402.363 GiB

License

MIT, following the source model. See LICENSE.