Doubao 1.5 Thinking Vision Pro 250428
ByteDance Seed · 2025-05-13
Doubao 1.5 Thinking Vision Pro 250428 is Volcano Engine's hosted snapshot of ByteDance Seed's Seed1.5-VL vision-language model. It combines a vision encoder reported at 532M parameters, an MLP adapter, and a Mixture-of-Experts language model described as approximately 20B active parameters. It accepts text, image, and video input over a 131,072-token context and returns text, supporting visual understanding, multimodal reasoning, document understanding, and GUI control. Deep thinking is enabled by default but can be disabled per request. Its VLM pre-training stages used stated budgets of 16B, 3T, and 240B tokens after separate vision-encoder and language-model pre-training, so exact cumulative training exposure is not disclosed.