| 1 | model_name: "nm-testing/Qwen3-Next-80B-A3B-Instruct-NVFP4" |
| 2 | accuracy_threshold: 0.75 |
| 3 | num_questions: 1319 |
| 4 | num_fewshot: 5 |
| 5 | server_args: >- |
| 6 | --enforce-eager |
| 7 | --max-model-len 4096 |
| 8 | --tensor-parallel-size 2 |
| 9 | --enable-expert-parallel |
| 10 | --speculative-config '{"method":"qwen3_next_mtp","num_speculative_tokens":1}' |
| 11 | --moe-backend=flashinfer_trtllm |