MAI-Image-2.5
#1microsoft · v2.5 · seit 2. Juni 2026 (Vollversion mit Image-Editing, Microsoft Build 2026); erste Text-zu-Bild-Vorabversion bereits 26. Mai 2026 · 23× · zuletzt 30. Juni 2026
MAI-Image-2.5 is Microsoft's in-house AI model for image generation and editing, developed by the MAI Superintelligence Team. It combines text-to-image generation with precise, localized image editing (e.g., object replacement, text correction, motion-blur removal) and emphasizes facial and identity preservation across edits. The model launched on June 2, 2026 at Microsoft Build 2026 and is available via Microsoft Foundry, the MAI Playground, and third-party platforms such as OpenRouter, Fireworks, and Baseten. A faster, cheaper variant, MAI-Image-2.5-Flash, launched alongside it.
Fonctionnalités
| Generation Time | No official figure found |
| License | Proprietary Microsoft licensing model; usage governed by Foundry terms of use |
| Max Resolution | Max. 1,048,576 pixels total (equivalent to 1024×1024); other aspect ratios possible as long as the pixel budget is maintained |
| Platform | Microsoft Foundry, MAI Playground, PowerPoint, OneDrive, OpenRouter, Fireworks, Baseten |
| Price | $5/1M text input tokens, $8/1M image input tokens, $47/1M image output tokens (Flash: $1.75/$1.75/$19.50-33) |
| Release Date | June 2, 2026 (2026-Launch build with image editing) |
