ERNIE 4.5 VL 28B A3B
ERNIE 4.5 VL 28B-A3B combines visual inputs with language in a sparse model. It produces text for image-based understanding tasks. This entry covers the post-trained checkpoint.
Model design and access
The chat model adds supervised and reinforcement-learning post-training. The card reports a 131,072-token checkpoint window. The published PyTorch weights are distinct from the smaller hosted endpoint limits listed by Qianfan.ERNIE-4.5-VL-28B-A3B-PT official model card
Practical use
Practical guidance. Use this model for image questions that require both recognition and explanation. Assess small text, charts and spatial relationships separately. If you resize or tile images, compare the extracted facts with the originals before judging the model’s reasoning.ERNIE-4.5-VL-28B-A3B-PT official model card
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | ERNIE-4.5-VL-28B-A3B-PT |
| Context window | 131,072 tokens Official checkpoint context window. Input and generation share this limit. ERNIE-4.5-VL-28B-A3B-PT official model card |
| Parameters | 28B total parameters; 3B active per tokenERNIE-4.5-VL-28B-A3B-PT official model card |
| Architecture | 28 layers; vision and language expert routingERNIE-4.5-VL-28B-A3B-PT official model card |
| License | Apache 2.0ERNIE-4.5-VL-28B-A3B-PT official model card |
Related models
- ERNIE 4.5 VL 28B A3B Base
Baidu · ERNIE 4.5
Sparse vision-language base checkpoint with 28 billion total parameters and three billion active parameters.
Launch date not confirmedDownloadable