sglang serve --model-path nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4 --tensor-parallel-size 2 --reasoning-parser nemotron_3 --tool-call-parser qwen3_coder --trust-remote-code --disable-radix-cache

python -m sglang.test.run_eval --base-url http://localhost:30000 --model nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4 --eval-name gsm8k --num-examples 1319 --repeat 1 --num-threads 512 --num-shots 5 --max-tokens 16000 --temperature 1.0 --top-p 0.95 --chat-template-kwargs '{"enable_thinking": true}'

ChatCompletionSampler initialized with self.system_message=None self.temperature=1.0 self.max_tokens=16000 self.reasoning_effort=None self.extra_body={'chat_template_kwargs': {'enable_thinking': True}}
100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1314/1314 [03:03<00:00,  7.15it/s]
Total latency: 183.893 s
Score: 0.969
[METRIC] gsm8k_score=0.9687975646879756 labels={"model": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4", "eval": "gsm8k"}
[METRIC] gsm8k_latency=183.893038809 labels={"model": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4", "eval": "gsm8k"}
Writing report to /tmp/gsm8k_nvidia_NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4.html
{'score:std': np.float64(0.17386443955744169), 'score': np.float64(0.9687975646879756)}
Writing results to /tmp/gsm8k_nvidia_NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4.json

sglang serve --model-path nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4 --tensor-parallel-size 2 --speculative-algorithm EAGLE --speculative-num-steps 3 --speculative-eagle-topk 1 --speculative-num-draft-tokens 4 --reasoning-parser nemotron_3 --tool-call-parser qwen3_coder --trust-remote-code --disable-radix-cache

python -m sglang.test.run_eval --base-url http://localhost:30000 --model nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4 --eval-name gsm8k --num-examples 1319 --repeat 1 --num-threads 48 --num-shots 5 --max-tokens 16000 --temperature 1.0 --top-p 0.95 --chat-template-kwargs '{"enable_thinking": true}'

ChatCompletionSampler initialized with self.system_message=None self.temperature=1.0 self.max_tokens=16000 self.reasoning_effort=None self.extra_body={'chat_template_kwargs': {'enable_thinking': True}}
100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1314/1314 [05:08<00:00,  4.25it/s]
Total latency: 308.988 s
Score: 0.971
[METRIC] gsm8k_score=0.9710806697108066 labels={"model": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4", "eval": "gsm8k"}
[METRIC] gsm8k_latency=308.98772795 labels={"model": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4", "eval": "gsm8k"}
Writing report to /tmp/gsm8k_nvidia_NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4.html
{'score:std': np.float64(0.1675798395536225), 'score': np.float64(0.9710806697108066)}
Writing results to /tmp/gsm8k_nvidia_NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4.json