Simat Labs
Book a demo
All research

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%