pytest -s -v tests/kernels/moe/test_gpt_oss_triton_kernels.py 

vllm serve openai/gpt-oss-20b --tensor-parallel-size 1 --max-num-seqs=16 

Writing report to /tmp/gpqa_openai__gpt-oss-20b-low_temp1.0_20251121_182105.html
{'chars': np.float64(66.06565656565657), 'chars:std': np.float64(235.89106420986758), 'score': np.float64(0.5681818181818182), 'score:std': np.float64(0.4953294254023493)}
Writing results to /tmp/gpqa_openai__gpt-oss-20b-low_temp1.0_20251121_182105.json
Writing all results to /tmp/gpqa_openai__gpt-oss-20b-low_temp1.0_20251121_182105_allresults.json
[{'eval_name': 'gpqa', 'model_name': 'openai__gpt-oss-20b-low_temp1.0_20251121_182105', 'metric': 0.5681818181818182}]

VLLM_ALL2ALL_BACKEND="deepep_high_throughput" vllm serve openai/gpt-oss-20b --tensor-parallel-size 1 --data-parallel-size 2 --enable-expert-parallel --no-enable-prefix-caching

Writing report to /tmp/gpqa_openai__gpt-oss-20b-low_temp1.0_20251121_184437.html
{'chars': np.float64(71.08207070707071), 'chars:std': np.float64(268.30453841690064), 'score': np.float64(0.5707070707070707), 'score:std': np.float64(0.4949752621616814)}
Writing results to /tmp/gpqa_openai__gpt-oss-20b-low_temp1.0_20251121_184437.json
Writing all results to /tmp/gpqa_openai__gpt-oss-20b-low_temp1.0_20251121_184437_allresults.json
[{'eval_name': 'gpqa', 'model_name': 'openai__gpt-oss-20b-low_temp1.0_20251121_184437', 'metric': 0.5707070707070707}]

vllm serve openai/gpt-oss-20b \
  --tensor-parallel-size 1 \
  --max-num-seqs 16

vllm bench serve \
  --model openai/gpt-oss-20b \
  --dataset-name sharegpt \
  --dataset-path ShareGPT_V3_unfiltered_cleaned_split.json \
  --max-concurrency 16 \
  --num-prompts 1000 \
  --num-warmups 60 \
  --ignore-eos

============ Serving Benchmark Result ============
Successful requests:                     1000      
Failed requests:                         0         
Maximum request concurrency:             16        
Benchmark duration (s):                  103.15    
Total input tokens:                      215312    
Total generated tokens:                  199033    
Request throughput (req/s):              9.69      
Output token throughput (tok/s):         1929.46   
Peak output token throughput (tok/s):    2104.00   
Peak concurrent requests:                33.00     
Total Token throughput (tok/s):          4016.74   
---------------Time to First Token----------------
Mean TTFT (ms):                          30.38     
Median TTFT (ms):                        20.28     
P99 TTFT (ms):                           579.15    
-----Time per Output Token (excl. 1st token)------
Mean TPOT (ms):                          8.03      
Median TPOT (ms):                        8.01      
P99 TPOT (ms):                           9.32      
---------------Inter-token Latency----------------
Mean ITL (ms):                           8.04      
Median ITL (ms):                         7.66      
P99 ITL (ms):                            17.10     
==================================================

============ Serving Benchmark Result ============
Successful requests:                     1000      
Failed requests:                         0         
Maximum request concurrency:             16        
Benchmark duration (s):                  96.91     
Total input tokens:                      215312    
Total generated tokens:                  199033    
Request throughput (req/s):              10.32     
Output token throughput (tok/s):         2053.88   
Peak output token throughput (tok/s):    2230.00   
Peak concurrent requests:                35.00     
Total Token throughput (tok/s):          4275.74   
---------------Time to First Token----------------
Mean TTFT (ms):                          19.89     
Median TTFT (ms):                        17.34     
P99 TTFT (ms):                           86.70     
-----Time per Output Token (excl. 1st token)------
Mean TPOT (ms):                          7.59      
Median TPOT (ms):                        7.55      
P99 TPOT (ms):                           8.64      
---------------Inter-token Latency----------------
Mean ITL (ms):                           7.59      
Median ITL (ms):                         7.37      
P99 ITL (ms):                            16.13     
==================================================