MAI-Image-2.5 Arena Rankings Vary by Category and Snapshot Date, Analysis Finds

Microsoft AI's launch materials claimed MAI-Image-2.5 ranked No. 2 in Arena's image-editing leaderboard and No. 3 in a separate text-to-image category. However, an Arena single-image editing leaderboard snapshot from August 7, 2026 placed the model at No. 4, behind GPT-Image-2 and Muse Image. The discrepancy highlights how benchmark rankings can shift over time and differ across leaderboard categories, making direct comparisons misleading without specifying the category and date. Analysts note that a text-to-image rank and an image-editing rank measure different capabilities and should not be treated as equivalent results. The case underscores the need for precise benchmark reporting, including the exact task category, snapshot date, and the competing models present at the time of measurement.