---

license: apache-2.0

base_model: sjakek/slimder-qwen38-reap384-depth36-mb2-4-5

pipeline_tag: text-generation

---

SLIMDER Qwen3.8 REAP-384 depth-36 Q4_K_M

Deployment GGUF for the promoted S3 depth-36 structural winner. The model has

36 transformer layers, retains 384 experts per layer, and removes original

macroblocks 2, 4, and 5.

Artifact integrity

• File: slimder-qwen38-reap384-depth36-mb2-4-5-Q4_K_M.gguf
• Size: 83,241,386,592 bytes
• SHA-256: e4bdd20d05ad6ea48b7e4700e70877514463abd4db16dc4d6ff3685a3fee4233
• Source revision: b3a1dc1e64854b4f69a984f4d3757d90b15bdf95
• llama.cpp revision: 9723942adc518b43c4b95dc4dce6906903eb5e09

The Hub LFS metadata was independently checked against the local size and

SHA-256 after upload.

Validation

The promoted BF16 model passed the repaired executable S3 functional gate 8/8.

This GGUF loaded and generated non-degenerate output under llama.cpp with a

2,048-token context and exited successfully. A short literal-output smoke did

not reach the requested exact string before its generation cap, so it is not

represented as an exact-string pass.

Full structural-screen, holdout, functional, materialization, and expert-map

evidence is in sjakek/slimder-qwen38-s3-results-20260831.

<!-- qwen38-perian-lineage:start -->

Qwen3.8 Perian project lineage

This repository is retained in the

Qwen3.8 Perian checkpoints collection.

Its exact position in the lineage is: Depth-pruning precursor at 36 layers. It retains 384 experts per layer and the full PLE table and predates the Perian QLoRA.

The final Qwen3.8 Perian GGUF release

combines three reductions and one post-training stage:

• depth: 48 to 32 transformer layers;
• routed-expert width: 384 to 288 experts per layer;
• PLE n-gram capacity: 320,001,446 to 160,000,768 rows (50%, about

25.60B parameters removed), using activation-aware bigram and

frequency-ranked trigram selections validated on a document-disjoint

5M-token holdout;

• rank-32 QLoRA on 12,558 normalized traces spanning math/STEM

reasoning, coding/debugging, agentic tool use, retrieval, and general

multi-step reasoning. The trace mixture draws from several frontier-model

families, including Fable 5, GLM 5.2, Kimi K3, Claude Opus 4.7,

Qwen3.8-Max, and GPT-5.6-Sol. The final merged milestone was trained through

9,336,692 supervised assistant tokens.

Earlier checkpoints in this collection do not inherit later stages merely by

being listed beside them; the stage statement above is authoritative for this

artifact.

<!-- qwen38-perian-lineage:end -->