Microsoft launches MAI-Image-2.6 AI image models in Foundry

Microsoft has released two new image generation models, MAI-Image-2.6 and the high-speed MAI-Image-2.6-Flash, via its Foundry developer platform. The standard model supports text-to-image generation and editing with resolutions up to 1.5K, while the Flash version is optimized for high-volume applications requiring faster response times.

Pricing for the standard model is set at approximately $5 per 1 million text tokens, $8 per 1 million image tokens, and $38 per 1 million image output tokens. The Flash variant offers reduced rates, including $1.75 for text tokens and $19 for image output tokens.

Reports indicate discrepancies in Microsoft’s official documentation regarding the models' performance metrics. While one Microsoft AI blog post claims a second-place ranking in Arena’s blind Text-to-Image comparisons, the Azure Foundry blog lists the model in third place for image editing. Additionally, conflicting figures were noted regarding the Flash model's efficiency and speed, with different posts citing varying percentages for resource reduction and different comparison benchmarks against OpenAI’s GPT-Image-2 models.