This is an experimental, structurally validated REAP expert-pruned checkpoint of
deepseek-ai/DeepSeek-V4-Flash-0731
. It keeps 160 of 256 routed experts in every MoE scope while preserving top-6 routing, shared experts, attention, embeddings, output head, mHC, compressor, and indexer tensors.
The checkpoint is intended for memory-constrained DeepSeek-V4 research and inference. It is not an official DeepSeek release, and pruning can reduce quality. Evaluate it on your own coding, reasoning, tool-use, and long-context workloads before production use.
What was pruned
Item
Value
Source revision
9e165c30e2704aec5d9d593cce3eebd58bbef1cb
Original routed experts
256 per MoE scope
Retained routed experts
160 per MoE scope
Routed experts removed
37.5%
Router top-k
6, unchanged
Backbone MoE layers
43
DSpark MTP blocks
3, mapped to backbone layers 40, 41, and 42
Safetensors shards
48
Tensor payload
107,803,320,952 bytes
Each retained expert is physically compacted into a contiguous 160-expert tensor. Learned router weights and biases are sliced to the same retained IDs. For the first three hash-routed layers, token-to-expert tables are remapped to the nearest retained current-revision router row by cosine similarity.
Observation provenance
This release transfers expert rankings from a prior DeepSeek-V4-Flash REAP observation run; it is
not
a fresh observation of the
0731
weights.
Coverage: coding, math, science, tool calling, and agentic tool trajectories; every expert was observed
Before transferring the ranking, all 43 old router matrices were aligned against the
0731
checkpoint. Every old expert row had the same expert ID as its nearest current-revision row. The weakest layer-mean same-index cosine was
0.9923683405
; the weakest individual row was
0.9460793734
. All three hash-routing tables were bit-identical. These results support identity transfer, but they do not make it equivalent to a fresh
0731
observation.
Validation
Structural validation passed with:
48/48 readable safetensors shards
45,821 tensors and exact index coverage
46/46 MoE scopes at K160
92 exact router-tensor comparisons
552 exact retained-expert old-to-new comparisons
12 exact protected-tensor comparisons
zero structural failures
The machine-readable build and validation records are included as
REAP_MANIFEST.json
,
reap_plan.json
, and files under
validation/
.
Runtime validation was performed on one NVIDIA DGX Spark / GB10 using the Anemll DeepSeek-V4 vLLM preview image, with CUDA graphs enabled. The bundled CUDA
sqrtsoftplus
router supports only a fixed set of expert counts, so K160 requires routing unsupported expert counts through vLLM's included pure-Torch router fallback. Expert execution remains on the B12X MXFP4 MoE backend.
The backbone-only API smoke passed at 8,192-token configured context:
/v1/models
became ready, the exact-answer probe returned
PRUNED_OK
in 6.19 seconds, and the coding probe returned
return a + b
in 1.60 seconds. Both stopped normally. The requests intentionally had no output-token cap. DSpark MTP was not enabled for this smoke.
Serving notes
Use a DeepSeek-V4-capable runtime. Generic Transformers releases may not understand the model's custom MXFP4, sparse-attention, tokenizer, and DSpark MTP components. The exact runtime smoke configuration and result are recorded under
validation/
.
For the tested vLLM preview lineage, patch
vllm_topk_softplus_sqrt
so expert counts outside the fused kernel's supported set
(16, 32, 64, 128, 192, 256, 320, 384, 512)
call the adjacent
_topk_softplus_sqrt_torch
implementation. Do not change the checkpoint back to 256 experts: that would invalidate the tensors and routing map.
Limitations
Expert importance came from a closely aligned earlier checkpoint, not a fresh
0731
observation.
Structural and smoke validation do not establish benchmark parity with the unpruned model.
This is an aggressive 37.5% routed-expert prune. Coding preservation is plausible because coding and agentic data were included, but must be measured rather than assumed.
The included three DSpark MTP blocks were pruned consistently; MTP runtime compatibility should be validated separately from backbone generation.
Credits and license
Thanks to
DeepSeek
for the base model,
Cerebras Research
for REAP, the
vLLM
contributors, and the Anemll/DGX Spark community for the DeepSeek-V4 Spark runtime work.
This derivative follows the source repository's MIT license. See
LICENSE
for the full text.
Runs of 0xSero DeepSeek-V4-Flash-0731-REAP on huggingface.co
4.1K
Total runs
0
24-hour runs
-309
3-day runs
-846
7-day runs
2.0K
30-day runs
More Information About DeepSeek-V4-Flash-0731-REAP huggingface.co Model
More DeepSeek-V4-Flash-0731-REAP license Visit here:
DeepSeek-V4-Flash-0731-REAP huggingface.co is an AI model on huggingface.co that provides DeepSeek-V4-Flash-0731-REAP's model effect (), which can be used instantly with this 0xSero DeepSeek-V4-Flash-0731-REAP model. huggingface.co supports a free trial of the DeepSeek-V4-Flash-0731-REAP model, and also provides paid use of the DeepSeek-V4-Flash-0731-REAP. Support call DeepSeek-V4-Flash-0731-REAP model through api, including Node.js, Python, http.
DeepSeek-V4-Flash-0731-REAP huggingface.co is an online trial and call api platform, which integrates DeepSeek-V4-Flash-0731-REAP's modeling effects, including api services, and provides a free online trial of DeepSeek-V4-Flash-0731-REAP, you can try DeepSeek-V4-Flash-0731-REAP online for free by clicking the link below.
0xSero DeepSeek-V4-Flash-0731-REAP online free url in huggingface.co:
DeepSeek-V4-Flash-0731-REAP is an open source model from GitHub that offers a free installation service, and any user can find DeepSeek-V4-Flash-0731-REAP on GitHub to install. At the same time, huggingface.co provides the effect of DeepSeek-V4-Flash-0731-REAP install, users can directly use DeepSeek-V4-Flash-0731-REAP installed effect in huggingface.co for debugging and trial. It also supports api for free installation.
DeepSeek-V4-Flash-0731-REAP install url in huggingface.co: