Qwen3-30B-A3B-Instruct-2507 GPTQ
Requirements
- vLLM : v0.11.1
- gptqmodel : v4.0.0
Performance
| id | Dataset | Metric | Samples | nv-score-bf16 | nv-score-fp8 | nv-score-w4a16 |
|---|---|---|---|---|---|---|
| 1 | aime25 | AveragePass@1 | 30 | 0.5666 | 0.594417 | 0.5 |
| 2 | gpqa_diamond | AveragePass@1 | 198 | 0.667 | 0.64848 | 0.6263 |
| 3 | mmlu_pro | AverageAccuracy | 1196 | 0.7751 | 0.7755 | 0.7466 |
| 4 | ifeval | prompt_level_strict_acc | 541 | 0.8355 | 0.83952 | 0.8429 |
| 5 | live_code_bench | Pass@1 | 1055 | 0.563 | 0.56358 | 0.491 |
- temperature 0.7
- top_p 0.8
- max_tokens 16384
- Downloads last month
- 11