GPT‑image‑2 Rollout

OpenAI’s new image model is moving into Microsoft Foundry, putting GPT-image-2 into Microsoft’s enterprise AI stack as of April 21. (techcommunity.microsoft.com) Microsoft said GPT-image-2 is now generally available and “rolling out today” in Microsoft Foundry, the company’s platform for building and governing AI apps and agents. Microsoft’s documentation describes Foundry as a service for developing, deploying, and managing AI systems at scale. (techcommunity.microsoft.com) (learn.microsoft.com) OpenAI introduced ChatGPT Images 2.0 the same day, saying the model improves text rendering, multilingual support, and visual reasoning. OpenAI’s help pages say the system can follow precise instructions, add text inside images, preserve fine details, and edit uploaded pictures. (openai.com) (help.openai.com) The rollout lands as Microsoft and OpenAI push image generation beyond consumer novelty and into production work such as marketing assets, design mockups, and localized visuals. OpenAI’s developer forum post said the model is built for images that are readable, on-brand, localized, and usable without heavy cleanup. (techcommunity.microsoft.com) (community.openai.com) Microsoft’s Azure documentation shows how quickly that shift is happening: DALL-E 3 was retired for new deployments on March 4, 2026, and customers are now directed to the gpt-image series instead. The same documentation says image generation in Foundry can be used through an image API or a broader responses API. (learn.microsoft.com 1) (learn.microsoft.com 2) The model’s stronger realism is also sharpening an old problem. Business Insider reported that OpenAI promoted ChatGPT Images 2.0 with fake screenshots and highly realistic generated pictures, underscoring how easily the tool can produce convincing false images. (businessinsider.com) Microsoft’s Foundry tools already pair image generation with governance features, and its agent documentation points users to Azure OpenAI transparency notes for limits and responsible AI guidance. That reflects the current enterprise pitch: generate faster, but log, label, and review what the system makes. (learn.microsoft.com 1) (learn.microsoft.com 2) OpenAI says ChatGPT Images 2.0 is available across all ChatGPT tiers, while “images with thinking” is limited to Plus, Pro, and Business and is still coming to Enterprise and Edu. The split suggests the consumer product and the enterprise rollout are moving in parallel, with Microsoft Foundry serving customers that want the model inside managed corporate workflows. (help.openai.com) (techcommunity.microsoft.com) For companies adopting it now, the appeal is simple: fewer prompt retries, cleaner text inside images, and easier deployment inside existing Microsoft systems. The tradeoff is just as clear: the better these images get, the more provenance and verification become part of the job. (openai.com) (businessinsider.com)