What is bysearch?
Our proprietary cross-modal large model that powers everything we build.
bysearch is our proprietary cross-modal large model. Where most models specialize in one modality, bysearch was designed from the ground up to unify understanding, expression, and measurement across text, images, audio, and video.
That design shows up in three capabilities. Multimodal understanding reads a page, a picture, and a voice note as one conversation. Multimodal generation creates text, images, and mixed content in response. And multimodal search retrieves the right answer across formats — finding the image that matches a sentence, or the passage that matches a frame of video. Because bysearch can measure what it understands, it can also verify: detecting generated content, spotting manipulation, and scoring trustworthiness.
Inside the PAI Workstation, bysearch anchors the AI copilot, real-time translation, and quality-assurance capabilities, working alongside the leading public models.
Why “bysearch”?
In Chinese mythology, Baize is the divine beast that knows all things — the legend behind the name of our model.
