Inference Efficiency Benchmark
20TTFT (ms)ACT
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| ACT2026.02 | 20 | — | 3.8 | — | — | |
| PaliGemma-3BHardware=NVIDIA A6000 GPU (48 GB)2026.04 | 60 | 64.3 | 6.02 | 10.7 | 1.554 | |
| DSPv22026.02 | 70 | — | 5.4 | — | — | |
| LLaVA-1.5-7BHardware=NVIDIA A6000 GPU (48 GB)2026.04 | 93 | 44.2 | 14.51 | 3 | 2.263 | |
| PaveGPT-7BHardware=NVIDIA A6000 GPU (48 GB)2026.04 | 236 | 39.7 | 16.81 | 2.4 | 2.518 | |
| LLaMA-3.2-11BHardware=NVIDIA A6000 GPU (48 GB)2026.04 | 253 | 30.8 | 22.31 | 1.4 | 3.247 | |
| InternVL-3.5-8BHardware=NVIDIA A6000 GPU (48 GB)2026.04 | 307 | 28.7 | 17.58 | 1.6 | 3.481 | |
| LLaVA-1.6-7BHardware=NVIDIA A6000 GPU (48 GB)2026.04 | 347 | 39 | 15.71 | 2.5 | 2.566 |