Downloads · 30 days
69
22% of all-time downloads
WaveCut/DeepSeek-V4-Flash-0731-REAM128-146B-exp-DS4-GGUF
DeepSeek-V4-Flash-0731-REAM128-146B-exp-DS4-GGUF is a machine learning model from WaveCut. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for gguf. The card lists the license as mit.
[!WARNING] Experimental — likely broken (renamed to -exp on 2026-08-12). This variant was produced by the cascaded REAM line (REAM160 → REAM128). Cascaded variants later showed severe multi-turn degradation: generatio…
Downloads · 30 days
69
22% of all-time downloads
All-time downloads
317
Public
Repo size
47.8 GB
Likes
0
Public
Click a slice to open those files.
.gguf47.6 GB · 100%
From the Hugging Face model README
[!WARNING] Experimental — likely broken (renamed to
-expon 2026-08-12). This variant was produced by the cascaded REAM line (REAM160 → REAM128). Cascaded variants later showed severe multi-turn degradation: generation typically collapses from the second request onward. No quality evidence was ever collected for this release. Kept for archival reference only; not recommended for use. A rebuilt line (REAM144 / REAM96, built in a single step from the base model) is in progress.
This single-file DS4 GGUF was produced directly from the committed REAM128 native checkpoint. It preserves the 43-layer, 128-routed-expert, one-shared-expert, top-k 6 topology.
Runtime requirement: this is a DS4-specific GGUF. Use the patched WaveCut DS4 fork with dynamic expert-count support. Generic llama.cpp and Transformers loading are not claimed.
6747e7718dd08f00b680d0c16231f2d59ec3747e.NATIVE_READY.json SHA-256: 8c32c76316574c75f18b45be69bd3b54c0c547531249b02f424a66485dc4820e.20260810T115212Z.| Artifact | Bytes | SHA-256 |
|---|---|---|
DeepSeek-V4-Flash-0731-REAM128-IQ2XXS-w2Q2K-AProjQ8-SExpQ8-OutQ8-chat-v2-imatrix.gguf | 47,553,155,968 | c209cf78b0fdf8fa0e164061346292299bd06b8c10355aba407dd3411adfd208 |
ream128-imatrix.dat | 225,448,838 | 3e01cbecd975a6b22d24a7f891871bfff1b27b2ca7f3e82749a17cf73fd1309e |
The GGUF contains 1,328 tensors. Its quantization template is pinned to antirez/deepseek-v4-gguf@591165d69ad5d27cf826dcc2cf687549e966f893.
The native checkpoint was produced by a training-free 160-to-128 expert merge using 3,072 calibration samples of 512 tokens each: 20% code, 10% math, 35% agent/tool, and 35% Russian/Cyrillic. Fresh layerwise saliency statistics, groups, merged experts, router mappings, and activations were computed for REAM128.
The importance matrix in this repository was generated fresh from the committed REAM128 native checkpoint; it was not reused from REAM160. DS4 consumed it in strict iMatrix mode. Conversion was direct from Safetensors to GGUF without an intermediate F16 checkpoint.
Build or install the patched DS4 fork at the pinned revision, then run:
./ds4 \
-m ./DeepSeek-V4-Flash-0731-REAM128-IQ2XXS-w2Q2K-AProjQ8-SExpQ8-OutQ8-chat-v2-imatrix.gguf \
--nothink
Remove --nothink only when intentionally enabling thinking with a suitable completion budget. Add backend, offload, context, and memory options for the installed CUDA or Metal build. The 47,553,155,968-byte model file alone is about 44.29 GiB; actual runtime memory also includes metadata, buffers, KV cache, and backend overhead.
The integrity checks prove artifact identity; they are not a quality or performance result.
MIT License, following the source checkpoint.