ERNIE 4.5 VL 424B A47B

A multimodal mixture-of-experts model from Baidu with 424B total parameters and 47B active per token, trained on both text and image data.

Strengths

Handles both text and image inputs with a large parameter base and efficient routing through mixture-of-experts architecture.

Best for

Tasks requiring vision and language understanding across a 123k token context, such as document analysis with images or visual reasoning at scale.

Limitations

As a large multimodal MoE model, it may have higher latency and memory requirements compared to smaller alternatives, and performance depends on the quality of image encoding.