gpt-image-1.5

Independent coverage on December 17, 2025 confirmed GPT Image 1.5's release to all ChatGPT users and via the API, citing up to 4x faster rendering and improved instructional understanding as headline improvements. The launch arrived shortly after OpenAI's Photoshop-ChatGPT integration announcement. Per the report, GPT Image 1.5 addresses a common pain point in AI image editing by keeping unrequested regions consistent rather than reinterpreting the whole image when users request targeted changes. OpenAI reportedly accelerated the release timeline in response to Google's Nano Banana Pro, sharpening the competitive context for this version.

Microsoft Foundry lists gpt-image-1.5 as an OpenAI model available "Direct from Azure," generally available and version-dated 2025-12-16. The catalog credits OpenAI as publisher and author, and frames the model as an efficient solution for diverse text and image tasks, including high-quality generation and editing scenarios hosted on Azure infrastructure. The Azure catalog enumerates concrete capabilities: text-to-image generation, image-to-image generation, inpainting, high-quality outputs up to 1024x1536 and 1536x1024 pixels, and face preservation. It also notes that gpt-image-1.5 includes all capabilities of gpt-image-1, while improving performance, quality, editing controls, and the ability to add or remove one image aspect while retaining others.

OpenAI's developer cookbook published a gpt-image-1.5 Prompting Guide on December 16, 2025, describing gpt-image-1.5 as their latest image generation model aimed at production-quality visuals and highly controllable creative workflows. The guide highlights high-fidelity photorealism, flexible quality–latency tradeoffs, robust facial and identity preservation across edits, reliable in-image text rendering, structured visuals like infographics and diagrams, precise style control, and strong real-world knowledge for accurate depictions of objects and scenes. The cookbook guide covers prompting fundamentals, recommending a consistent prompt structure (background and scene, then subject, then key details, then constraints) to maximize output fidelity. Although the page is labeled archived and may reference outdated models or APIs, its capability description remains first-party from OpenAI authors Annika Brundyn, Mandeep Singh, and Emre Okcular, serving as an authoritative reference for gpt-image-1.5 features and prompting patterns.

OpenAI's official model catalog now lists gpt-image-1.5 as Deprecated, describing it as "our previous image generation model" with "better instruction following and adherence to prompts" than its successor. The page documents multimodal text and image input and output, with audio and video unsupported, and lists the model as serving Chat Completions, Responses, Realtime, Batch, and Assistants endpoints via the v1 API. The deprecated model carries $5.00 per million input text tokens, $10.00 per million output tokens, and $8.00/$32.00 per million image input/output tokens, with per-image prices ranging from $0.009 at low quality to $0.20 at high quality for 1024x1024 up to 1536x1024 outputs. Developers can still access the model through OpenAI's API catalog and a linked usage guide for migration planning.

OpenAI announced GPT Image 1.5 on December 16, 2025 as its new flagship image generation model, releasing it simultaneously in the ChatGPT Images experience and via the API under the identifier gpt-image-1.5. Business and Enterprise ChatGPT access was slated to follow the initial rollout. According to OpenAI's launch post, the model generates images up to 4x faster than its predecessor and handles precise edits more reliably, preserving lighting, composition, and subject likeness across outputs and successive edits. It supports adding, subtracting, combining, blending, and transposing operations while keeping unchanged elements intact.

An Interesting Engineering piece syndicated via MSN corroborates OpenAI's December 2025 release of GPT-Image 1.5, noting immediate availability to all ChatGPT users worldwide and through OpenAI's API under the name GPT-Image-1.5. The article reports the model runs up to four times faster than its predecessor, follows i The piece frames the release as a competitive response to Google's Nano Banana Pro and quotes Fidji Simo, CEO of applications at OpenAI, on the shift toward a visual-first workflow inside ChatGPT. A dedicated image creation space with preset styles and trending designs was introduced alongside the model. As of the Sept