Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

LLM Inference on Qwen3-8B synthetic workload (mu_L=512, mu_O=256)

48,015Throughput (tok/s)

EB+

6,949.5617,610.7828,27238,933.22May 30, 2026
Updated 1d ago

Evaluation Results

MethodLinks
2026.05
48,0156.3399.71
2026.05
46,0427.60197.44
2026.05
45,7310.78829.18
2026.05
45,5390.77929.44
2026.05
39,6493.60822.81
2026.05
26,8570.047.03
2026.05
26,5450.0427.11
2026.05
22,8250.0918.08
2026.05
21,7101.958.4
2026.05
21,5261.958.9
2026.05
19,92645.363.8
2026.05
16,4676.21958.64
2026.05
16,22845.9150
2026.05
8,6910.08422.1
2026.05
8,6060.16421.1
2026.05
8,5290.08722.3