Qwen3-VL-8B-Instruct

Original model repository: Qwen/Qwen3-VL-8B-Instruct

Model Introduction

Qwen3-VL-8B-Instruct is an instruction-tuned Vision-Language Model (VLM) for understanding images, videos, and text. It combines a vision encoder with a dense autoregressive language model. The model supports visual question answering, multilingual OCR, document understanding, visual grounding, spatial reasoning, video understanding, visual coding, and visual-agent tasks.

Deployment Metrics

Model Parameters

Metric	Value
Total model parameters	8.767B
Vision model (ViT) parameters	576.4M
Language model (LM) parameters	8.191B

Parameter counts are calculated from the tensors stored in the upstream checkpoint.

Performance Metrics

Test Configuration

Metric	Matrix6P W8A8	S6P W8A8	S6P W4A8
ViT image size	448 × 448	448 × 448	448 × 448
Sequence length	512	512	512
Maximum context length	1024	1024	1024
BPU cores (ViT / Prefill / Decode)	4 / 4 / 4	4 / 4 / 4	4 / 4 / 4

Performance Results

Metric	Matrix6P W8A8	S6P W8A8	S6P W4A8
ViT latency	28.593 ms	25.396 ms	25.111 ms
Time to first token (TTFT)	193.200 ms	164.228 ms	159.716 ms
Prefill throughput	3,211.139 tokens/s	3,807.482 tokens/s	3,928.261 tokens/s
Decode throughput	19.840 tokens/s	20.613 tokens/s	31.789 tokens/s

Memory Usage

Metric	Matrix6P W8A8	S6P W8A8	S6P W4A8
BPU memory	9.4 GB	9.6 GB	6.4 GB
CPU memory	2.5 GB	2.5 GB	2.5 GB

Note: TTFT includes preprocessing and ViT latency. Memory values represent the peak memory usage measured during the specified performance test.