ByteDance / Seedream-3 (Seedream 3.0)

Seedream 3.0 is a bilingual (Chinese & English) text-to-image foundation model by ByteDance. Building on Seedream 2.0, it improves native 2K resolution output (without upscaling), faster inference, better prompt understanding and alignment, and especially good performance in scenarios with text/signage/layout. It’s well suited for users needing both speed and visual quality, including those who mix Chinese/English prompts.