Artificial Analysis says GPT Image 2.5 Flare leads in 5 of its 9 text-to-image capabilities

Artificial Analysis reports that GPT Image 2.5 Flare leads in five of the nine capabilities measured in its text-to-image arena. Its largest leads over the next-best models are in Knowledge, Material and Physics—covering real-world facts, surface properties and physical behavior. It also narrowly leads Complex Compositions and Text Rendering. In preliminary speed tests, the group says Flare (max) reduced median generation time by 63% versus GPT Image 2 (high), from 142.9 to 52.7 seconds. Those tests used 1024 × 1024 PNG outputs, with timing covering both inference and image transfer across several runs.