Distillation
Same answer.About 30× sooner.
We distilled Qwen3-VL-8B into a small model fine-tuned for a computer vision task. It matches the 8B model's accuracy and returns each answer in about 16 ms.
Latency per video window, batch 1, one A100
- Qwen3-VL-8Breads every frame
- 513 ms
- Simat Labs distilled modelsame accuracy
- 16 ms
- faster than Qwen3-VL-8B on computer vision tasks, at the same accuracy
- ~30×
- throughput on Qwen3-VL-32B from our 4-bit quantization
- +59%