Artificial Analysis 的 Image Arena 评测显示，MAI-Image-2.6-Flash 在图像编辑中表现最强的三类操作是推理式编辑、构图与取景、身份保持编辑。相比 MAI-Image-2.5-Flash，它在除增强与修复外的所有编辑操作上都缩小了与前沿水平的差距，其中文本/符号编辑、构图与取景、身份保持编辑提升最大。
Where MAI-Image-2.6-Flash is strongest in Image Editing: Reasoning-Based Edits, Composition & Framing, and Identity Preserving Edits
Our Image Arena taxonomy measures 7 editing actions and 10 use cases for Image Editing, each with its own leaderboard. Editing actions measure what an edit asks the model to do.
MAI-Image-2.6-Flash is closest to the category frontier in Reasoning-Based Edits, followed by Composition & Framing and Identity Preserving Edits. Against MAI-Image-2.5-Flash it closes the gap to the frontier on every editing action except Enhancement & Restoration, with the largest gains in Text or Symbol Edits, Composition & Framing, and Identity Preserving Edits.
➤ Reasoning-Based Edit covers edits the prompt implies rather than states, where the model has to work out what must change.
➤ Composition & Framing measures 2D and 3D reframing, including camera angle changes, zoom-outs that extend a scene, and rearranging graphic layouts while preserving design, spacing, and style.
➤ Identity Preserving Edit covers changes to people, faces, and characters that must keep identity intact.
来源：@ArtificialAnlys · x.com