---

license: mit

base_model: deepseek-ai/DeepSeek-V4.1-Flash

tags:

• gguf
• deepseek
• lucebox
• rocmfp

---

DeepSeek V4.1 Flash for the Lucebox engine (ROCMFP block formats)

Lucebox builds of DeepSeek V4.1 Flash for the Lucebox DS4.1 engine

(AMD R9700 + Strix Halo). Routed experts in the engine's ROCMFP block

formats; attention, shared experts and head in Q8_0; both Engram tables

embedded. Calibrated with the published 262,144-token imatrix

(smalinin/DeepSeek-V4.1-Flash-GGUF, converted in imatrix/).

| file | routed experts | expert bytes | KL top-256 vs teacher | top-1 | ppl |

|---|---|---|---|---|---|

| teacher (fp4 checkpoint) | | 269 GiB | 0 | 100% | 8.45 |

| antirez Q2 (IQ2_XXS / Q2_K) | | 142 GiB | 0.260 | 78.3% | 9.71 |

| DeepSeek-V4.1-Flash-ROCMFP23.gguf | fp2 gate/up, fp3 down | 179 GiB | 0.272 | 77.9% | 9.80 |

| DeepSeek-V4.1-Flash-ROCMFP2S.gguf | fp2s everywhere | 158 GiB | 0.266 | 77.7% | 9.59 |

| DeepSeek-V4.1-Flash-ROCMFP2S-MIX10.gguf | fp2s, fp3 down on layers 0-9 | 164 GiB | 0.246 | 78.3% | 9.45 |

Formats: Q2_0_ROCMFP2 (ggml type 107, 10 bytes per 32 weights, 2.5 bits,

codebook {-1, 0, 1, 2} x a ue4m3 scale per 16 weights), Q3_0_ROCMFPX

(type 104, 14 bytes per 32, 3.5 bits). fp2s files use type 107 with bit

7 of a half-block's scale byte meaning "mirror the codebook"; a reader that

ignores that bit produces garbage, so use an engine build that honors it.

KL is measured on 8,184 held-out fineweb-edu tokens against the unquantized

model through the same reference forward pass (eval/ holds the teacher

log-probs and the tokens). Recipe, converter patches and harness:

Lucebox repo, lucebox_training/quantization.

Image input

DeepSeek-V4.1-Flash-mmproj-BF16.gguf (0.90 GiB): this checkpoint's vision

encoder (32-layer ViT and the aligner into the 5120-wide model), its 266

tensors copied byte for byte in BF16 by the engine's

server/tools/export_ds4v_mmproj.py, with the checkpoint revision recorded

(sha256 in vision/). Pass it with --mmproj to an engine build with V4.1

image input, next to these September files only: the October 17 checkpoint

retrained its encoder and ships its own in

Lucebox/DeepSeek-V4.1-Flash-1017-ROCMFP-GGUF. The engine refuses an encoder

from another checkpoint than the target's.

IQ2 / IQ3 + MXFP8 builds

ds41-lucebox-final.gguf (recommended) and ds41-lucebox-minus10gib.gguf (10 GiB smaller): routed experts in

IQ2_XXS / IQ3_XXS, dense projections in MXFP8 (type 117, needs an engine with MXFP8, see Luce-Org/lucebox#800).

Correct answers on our internal hard bench; exact = plain greedy, no drafter.

| internal hard bench | final, Lucebox profile + DSpark | final, exact | minus10gib, exact | antirez Q2 |

|---|---|---|---|---|

| file size | 366.1 GB | 366.1 GB | 355.4 GB | 365.7 GB |

| 96 questions (8K tokens max) | 94 | 93 | 91 | 86 |

| 32 long-reasoning questions (16K tokens max) | 27 | 30 | 30 | 26 |

| 49 long-reasoning questions (16K tokens max) | 37 | 41 | 34 | 29 |

| total (177 questions) | 158 | 164 | 155 | 141 |