| base_model: Qwen/Qwen3-1.7B | |
| [EXL3](https://github.com/turboderp-org/exllamav3) quantization of [Qwen3-1.7B](https://huggingface.co/Qwen/Qwen3-1.7B), 8 bits per weight, including output layers. | |
| ### HumanEval (argmax) | |
| | Model | Q4 | Q6 | Q8 | FP16 | | |
| | ------------------------------------------------------------------------------------------ | ---- | ----- | ----- | ----- | | |
| | [Qwen3-1.7B-exl3-8bpw-h8](https://huggingface.co/isogen/Qwen3-1.7B-exl3-8bpw-h8) | 0.0% | 70.7% | 68.3% | 68.9% | | |
| | [Qwen3-1.7B-Base-exl3-8bpw-h8](https://huggingface.co/isogen/Qwen3-1.7B-Base-exl3-8bpw-h8) | 0.0% | 66.5% | 70.7% | 70.1% | | |