正文
Re GPT Image 2.5 Flare leads in 5 of the 9 Text to Image capabilities we measure, with the largest leads in Knowledge, Material, and Physics.
Our Text to Image arena scores models on two axes: nine capabilities, which are the skills that go into making an image, and ten use cases, which measure how those capabilities combine in real workloads.
Flare's largest leads over the next-best model are in Knowledge and Physics, both over GPT Image 2 (high), and Material, over Microsoft's MAI-Image-2.6. Flare also narrowly leads Complex Compositions and Text Rendering.
➤ Knowledge covers real landmarks, species, and domain facts across science and common sense.
➤ Material covers surface properties, transparency, subsurface scattering, and texture realism.
➤ Physics covers gravity, support, collision, and thermal and state change.
Our Text to Image arena scores models on two axes: nine capabilities, which are the skills that go into making an image, and ten use cases, which measure how those capabilities combine in real workloads.
Flare's largest leads over the next-best model are in Knowledge and Physics, both over GPT Image 2 (high), and Material, over Microsoft's MAI-Image-2.6. Flare also narrowly leads Complex Compositions and Text Rendering.
➤ Knowledge covers real landmarks, species, and domain facts across science and common sense.
➤ Material covers surface properties, transparency, subsurface scattering, and texture realism.
➤ Physics covers gravity, support, collision, and thermal and state change.