# AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration

*Source* · `src-lin-2023-awq` · status: **n/a**

| field | value |
|---|---|
| kind | Peer-reviewed paper |
| authors | Ji Lin, Jiaming Tang, Haotian Tang, Shang Yang, Wei-Ming Chen, Wei-Chen Wang, Guangxuan Xiao, Xingyu Dang, Chuang Gan, Song Han |
| venue | MLSys 2024 (Best Paper Award) |
| date | 2023-06-01 |
| url | https://arxiv.org/abs/2306.00978 |
| arxiv | 2306.00978 |
| accessed | 2026-10-24 |

## Notes

4-bit weight-only quantization protecting about 1% salient weights chosen from activation statistics; TinyChat runtime over 3x faster than Hugging Face FP16 on desktop and mobile GPUs.

## Backlinks (computed)

**cited by:** [On-device small language models for characters](../technique/on-device-language-models.md)

## Citation

Ji Lin, Jiaming Tang, Haotian Tang, Shang Yang, Wei-Ming Chen, Wei-Chen Wang, Guangxuan Xiao, Xingyu Dang, Chuang Gan, Song Han (2023-06-01). *AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration*. MLSys 2024 (Best Paper Award) https://arxiv.org/abs/2306.00978

---
Provenance: created in run `2026-10-24-techniques-llm-agents`.