Inference Latency Measurement on H100 GPU (16k inputs, test)
0.052Latency (s)Lemon
Evaluation results
| Method | Links | |
|---|---|---|
| LemonBackbone Size=7B2025.12 | 0.052 | |
| Qwen2.5-VLBackbone Size=7B2025.12 | 0.0588 | |
| LLaVA-1.5Backbone Size=13B2025.12 | 0.0672 | |
| ShapeLLMBackbone Size=7B2025.12 | 0.0745 | |
| 3D-LLMBackbone Size=7B2025.12 | 0.0762 |