---
license: apache-2.0
base_model: Qwen/Qwen3.5-35B-A3B
library_name: transformers
tags:
- dashq
- quantized
- post-training-quantization
---

# Qwen3.5-35B-A3B-DASHQ-INT3-g128

This repository contains a DASH-Q packed quantized checkpoint for `Qwen/Qwen3.5-35B-A3B`.

DASH-Q checkpoints require the lightweight DASH-Q runtime package for loading. They are not plain Transformers checkpoints because linear layers are stored as `PackedQuantizedLinear` modules.

## Install

```bash
pip install git+https://github.com/JaeminK/dashq.git
```

## Load

```python
from dashq import load_quantized

model, tokenizer = load_quantized(
    "jkim96/Qwen3.5-35B-A3B-DASHQ-INT3-g128",
    device_map="auto",
)
```

## Quantization

| Field | Value |
| --- | --- |
| Base model | `Qwen/Qwen3.5-35B-A3B` |
| Bits | `3` |
| Group size | `128` |
| Scale/zero dtype | `float16` |
| Calibration dataset | `wikitext2` |
| Calibration samples | `128` |
| Sequence length | `2048` |
| Original size | `71.9039 GB` |
| Quantized size | `17.4800 GB` |

## Evaluation

| Metric | Value |
| --- | ---: |
| `wikitext2_ppl` | 7.1423 |
| `zero-shot accuracy avg` | 69.9820 |
| `arc_challenge` | 60.2389 |
| `arc_easy` | 81.2710 |
| `commonsense_qa` | 84.4390 |
| `hellaswag` | 80.4720 |
| `lambada_openai` | 69.3771 |
| `openbookqa` | 43.2000 |
| `piqa` | 82.2633 |
| `truthfulqa_mc2` | 55.1749 |
| `winogrande` | 73.4017 |