OpenAI's New GPT-image-2 Model Is Now Available in Microsoft Foundry

Yesterday, OpenAI launched ChatGPT Images 2.0, it’s new image generation model. Now, Microsoft has started rolling out OpenAI’s GPT-image-2 inside its Foundry platform, marking the model’s general availability for enterprise use.

GPT-image-2 now available in Microsoft Foundry

The rollout means businesses can start using GPT-image-2 for image generation tasks. Microsoft positions it as a production-ready model, aimed at teams building design assets at scale. For those unaware, the new image AI model now supports multiple output sizes, including 4K resolution, with a defined pixel budget cap of 8,294,400 pixels.

However, if requests exceed that limit, the system automatically resizes them to fit within constraints. Supported formats include 1024×1024, 1536×1024, and 1024×1536, with dimension rules requiring values to be multiples of 16.

Microsoft is also releasing an intelligent routing layer with two modes. One sticks to legacy size tiers like small, standard, and large outputs, while the other uses token-based buckets for more flexible generation. The idea is to let the system handle optimization without requiring manual input.

Pricing structure targets enterprise usage

On the pricing side, GPT-image-2 follows a token-based model. For image generation, input tokens are priced at $8 per million, cached inputs at $2, and output tokens at $30. For text-related usage within the same model, input tokens cost $5 per million, cached inputs $1.25, and output tokens $10.

These rates place GPT-image-2 firmly in the enterprise tier, especially for teams generating large volumes of visual content. Just in case you are wondering what the new AI image model can pull off, here are a few examples:

Have you already tried ChatGPT Images 2.0? If yes, let us know what you think about the new image model.