OpenAI: GPT Image 2

GPT Image 2 is an image generation model from OpenAI, listed on api14 under the model id openai/gpt-image-2. It sits in the image category, and OpenAI describes it as its latest image generation model, supporting high fidelity image generation and editing. If your application needs to produce or modify images rather than text, this is the model family to look at first.

The model accepts text and image inputs and returns images as its only output modality. That combination is what makes editing possible: you can send a prompt on its own, or pair a prompt with reference images to guide the result. The context window is 400,000 tokens, with a maximum output of 360,000 tokens. Generation is controlled through a set of image parameters. Aspect ratio can be set to 1:1, 3:2, 2:3, 4:3, 3:4, 16:9, 9:16, 21:9 or auto. Quality accepts auto, low, medium or high. Background accepts auto or opaque. You can request between 1 and 10 images per call with n, supply between 0 and 16 input references, and set output compression anywhere from 0 to 100.

Those capabilities point to a fairly wide set of practical uses. Text to image generation covers marketing assets, app and web artwork, concept exploration and placeholder or production imagery. Because the model also takes image inputs, editing workflows fit naturally: restyling an existing asset, adjusting a composition, or producing variations that stay close to a supplied reference. The input_references parameter, which allows up to 16 images, suits tasks where several source images need to inform a single result, such as combining elements from multiple references. The n parameter helps when you want several candidates in one request, and the aspect ratio and quality settings let you match output to a specific slot in a layout, from square social formats to wide 21:9 banners.

You call the model through api14's OpenAI-compatible API using the model id openai/gpt-image-2, so existing OpenAI client libraries and request shapes generally carry over. The supported parameters are max_tokens, response_format, seed and structured_outputs, alongside the image parameters described above: aspect_ratio, quality, background, n, input_references and output_compression. The seed parameter is useful when you need reproducible results across runs, and response_format lets you control how the image data comes back. Because the interface is OpenAI-compatible, switching to this model from another OpenAI image model is mostly a matter of changing the model id and adjusting parameters.

One thing to keep in mind is the output modality. GPT Image 2 returns images only, so it is not the right choice when you need text completions, chat responses, or any other text output; a text model fits those cases better. Within image work, the parameter ranges define the boundaries: at most 10 images per request, at most 16 input references, and quality and compression values limited to the listed options. If your task falls inside those bounds, this model covers both generation and editing through a single endpoint.