DeepSeek-V4.1-Flash-1017-Sushi-STQ-1.75bpw

deepseek-ai/DeepSeek-V4.1-Flash-1017 packed for sushi, a native inference engine for Apple Silicon.

part	how it is stored
routed experts (40 layers x 384)	STQ (Sushi Trellis Quant) with the PAIR1 2-D codewords, 1.75 average bits
lm_head	STQ with the SPEC1 codebook table, 6 bits
embedding	4-bit affine
Engram tables	as released (FP8 E4M3 with E8M0 scales), read from the SSD and never loaded into GPU memory
DSpark (the 3 speculative-decoding stages)	included: stage experts STQ PAIR1 at 2 bits, stage trunk STQ SPEC1 at 4 bits
vision tower	included, bf16: image input
context	up to 1,048,576 tokens
disk	~309 GiB: 120 GiB of weights + 189 GiB of Engram tables

Sushi Engine Only. This pack needs sushi v1.4.0 or later on Apple Silicon with macOS 26.2 or later.

How to run it (install, launch flags, memory and context per Mac): sushi usage.

Quality

KLD against the original checkpoint, run by sushi with no lossy step of its own (FP8 trunk and FP4 experts expanded, full-width KV, DSpark off). Lower is better.

reading	KLD	top-1 agreement
16 prompts x 512 tokens, scored to the first EOS	0.2197	85.64%
eight categories (code, math, reasoning, design, wikitext, multilingual, story, science), 16 x 512	0.1843	88.53%

On the same fixtures and teacher, oQ2.5e (oMLX, 2.96 bits per weight, 342.7 GB on disk, its logits taken through mlx-serve's DeepSeek path) reads 0.2575 (85.33%) on the 16 x 512 reading and 0.1717 (89.17%) on the eight categories. This pack is 332.1 GB on disk.

KLD by category, to the first EOS:

pack	code	math	reasoning	design	wikitext	multilingual	story	science
Sushi-STQ-1.75bpw	0.0357	0.0598	0.1176	0.1593	0.2714	0.2733	0.4878	0.0985
oQ2.5e	0.0298	0.0677	0.0831	0.1474	0.3014	0.2868	0.4152	0.0805

The two readings use different prompts and are not comparable with each other.

Tool calling

tool-eval-bench 3.0.0 in hard mode (seed 42, no live tools), served by sushi 1.4.0 with its defaults (thinking at high, DSpark on) at --ssd-budget-gb 108: 92 / 100, 178 of 194 points over 97 scenarios.

Credits and license

• DeepSeek-AI: the base model, deepseek-ai/DeepSeek-V4.1-Flash-1017. This pack is a derivative under its MIT License (LICENSE).
• ddalcu: mlx-serve. sushi is a fork of it.