---
license: apache-2.0
base_model: Qwen/Qwen3.6-35B-A3B
tags:
  - moe
  - expert-pruning
  - code
  - gguf
  - llama.cpp
pipeline_tag: text-generation
---

# Qwen3.6-35B-A3B coding specialist, 25% experts kept, healed (GGUF)

Companion checkpoint for [*Half the Experts, All the Code: One-Shot Domain
Pruning of Mixture-of-Experts LLMs for Coding*](https://arxiv.org/abs/2607.16721)
(arXiv:2607.16721). Pipeline and the exact keep-set JSON that produced it:
[github.com/anik-jha/moep](https://github.com/anik-jha/moep).

75% of the routed experts were removed from Qwen3.6-35B-A3B under the REAP
criterion calibrated on a coding-weighted mix, then the survivors were healed by
distilling from the base model with the expert tensors frozen (LoRA on attention,
router, and shared expert only).

## Read this before using it

This is a research artifact, and the paper's headline finding is a negative one:
**at this memory budget you are better off quantizing the full model than pruning
it.** The healed checkpoint below sits at 11.2 GB and scores 0.841 on HumanEval+.
A 2-bit IQ2_M quantization of the *unpruned* model sits at 12.1 GB and scores
0.896. Pruning to 25% and healing does not close that gap; five separate attempts
to close it are reported in the paper and all failed.

The checkpoint is released because healing is GPU-nondeterministic and therefore
not reproducible from the keep-set JSON, unlike every other point in the study.
It is the artifact behind the paper's healing result, not a recommended
deployment. If you want something to actually run, use the 50%-keep specialist:
[Qwen3.6-35B-A3B-coding-reap50-GGUF](https://huggingface.co/anik-jha/Qwen3.6-35B-A3B-coding-reap50-GGUF).

## Evaluations

All numbers from `runs/qwen36/eval.json` in the pipeline repo, greedy decoding,
EvalPlus HumanEval+ and MBPP+ (pass@1). Only the **Q8_0** file was functionally
evaluated.

| Checkpoint | Size | HumanEval+ | MBPP+ |
|---|---|---|---|
| Base Qwen3.6-35B-A3B, Q8_0 | 36 GB | 0.890 | 0.772 |
| Base, full model, IQ2_M (iso-memory comparator) | 12.1 GB | 0.896 | 0.730 |
| 25% keep, REAP, no healing, Q8_0 | 11.2 GB | 0.787 | 0.659 |
| **25% keep, REAP, healed, Q8_0** (this repo) | **11.2 GB** | **0.841** | **0.685** |

Healing recovers 54 of the 103 HumanEval+ points lost to pruning (0.787 → 0.841
against a base of 0.890) and 26 of 113 on MBPP+. Held-out perplexity improves
from 5.45 to 4.85 on code and from 30.9 to 9.4 on general text.

**The `Q4_K_M` file in this repo has not been functionally evaluated.** It passed
an export smoke test and was benchmarked for throughput, nothing more. Do not
attribute the table above to it.

## Files

| File | Size | Notes |
|---|---|---|
| `qwen36-reap25-healed-Q8_0.gguf` | 11.2 GB | evaluated, numbers above |
| `qwen36-reap25-healed-Q4_K_M.gguf` | 6.5 GB | smoke-tested only, not evaluated |

Throughput measured with `llama-bench` on one DGX Spark (GB10, 119 GB unified
memory): Q8_0 60.7 tok/s generation, 2780 tok/s prefill; Q4_K_M 82.0 tok/s
generation, 3081 tok/s prefill.

## License

Weights are a derivative of Qwen3.6-35B-A3B and inherit its Apache-2.0 license.
The pipeline that produced them is MIT.