InternVL3 5 GPT OSS 20B A4B Preview is a vision-language model that punches above its active parameter count — with only 4 billion parameters active at inference time out of 20 billion total, it uses a mixture-of-experts architecture to stay efficient. It handles image understanding tasks like chart reading, document parsing, and visual question answering with reasonable fluency. As a preview release, expect rough edges and evolving behavior.