LLM Inference on Qwen3 Google Pixel 9 Pro XL 0.6B (test)
591.01Prefill Throughput (min, tokens/sec)ET Vulkan
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| ET VulkanHardware=GPU, GS=1282026.05 | 591.01 | 601.83 | 20.71 | 21.12 | 337 | |
| ET VulkanHardware=GPU, GS=322026.05 | 313.1 | 329.99 | 20.59 | 20.81 | 457 | |
| llama.cppHardware=CPU, GS=322026.05 | 240.6 | 241.1 | 46.5 | 46.6 | 442 | |
| ONNXHardware=CPU, GS=1282026.05 | 229.85 | 285.7 | 32.71 | 34.27 | 323 | |
| ET XNNPACKHardware=CPU, GS=1282026.05 | 187.99 | 379.51 | 41.92 | 42.22 | 376 | |
| ET XNNPACKHardware=CPU, GS=322026.05 | 169.13 | 299.54 | 37.07 | 40.38 | 417 | |
| ONNXHardware=CPU, GS=322026.05 | 144.42 | 259.93 | 26.52 | 35.23 | 376 | |
| LiteRTHardware=CPU, GS=1282026.05 | 101.14 | 109.68 | 10.22 | 10.31 | 299 | |
| LiteRTHardware=CPU, GS=322026.05 | 96.03 | 97.74 | 9.53 | 10.08 | 326 | |
| llama.cppHardware=GPU, GS=322026.05 | 75.5 | 76.4 | 24 | 31.3 | 442 |