| --- |
| license: other |
| tags: |
| - heal |
| - horizon |
| --- |
| |
| # Qwen3-VL-2B-Instruct |
|
|
| **Original model repository:** [Qwen/Qwen3-VL-2B-Instruct](https://huggingface.co/Qwen/Qwen3-VL-2B-Instruct) |
|
|
| ## Model Introduction |
|
|
| Qwen3-VL-2B-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 | 2.128B | |
| | Vision model (ViT) parameters | 407.0M | |
| | Language model (LM) parameters | 1.721B | |
|
|
| Parameter counts are calculated from the tensors stored in the upstream checkpoint. |
|
|
| ### Performance Metrics |
|
|
| | Chips | Scenario | Data Type | ViT Image Size | Sequence Length (tokens) | Maximum Context Length (tokens) | BPU Cores (ViT / Prefill / Decode) | ViT Latency (ms) | TTFT (ms) | Prefill TPS (token/s) | Decode TPS (token/s) | BPU Memory (GB) | CPU Memory (GB) | |
| |---|---|---|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:| |
| | J6P | Image | W8A8 | 448 × 448 | 512 | 1024 | 4 / 4 / 4 | 18.489 | 76.787 | 9,567.264 | 61.967 | 3.0 | 1.3 | |
| | S6P | Image | W8A8 | 448 × 448 | 512 | 1024 | 4 / 4 / 4 | 16.662 | 68.262 | 10,786.589 | 67.446 | 3.1 | 1.3 | |
| | S6P | Video | W8A8 | 448 × 448 per frame | 1024 | 4096 | 4 / 4 / 4 | 80.971 | 452.232 | 7,462.987 | 43.671 | 4.9 | 1.4 | |
|
|
| > **Note:** TTFT includes preprocessing and ViT latency. Memory values represent the peak memory usage measured during the specified performance test. |
|
|
