OpenAI: GPT-5 Image

GPT-5 Image is a model from OpenAI that combines the GPT-5 language model with image generation capabilities. It sits in the image category while also handling text, so it is a multimodal model rather than a pure image generator. OpenAI released it on 14 October 2025. If you already use GPT-5 for text work, this model extends the same family into image generation and editing.

The model accepts text, images and files as input, and it can produce both text and images as output. Its context window is 400,000 tokens, with a maximum output of 128,000 tokens. On the image side it exposes a set of generation controls: aspect ratio (1:1, 3:2, 2:3 or auto), quality (auto, low, medium or high), background (auto, transparent or opaque), a count n between 1 and 10, up to 16 input references, and output compression from 0 to 100. The presence of input references means you can supply images for the model to work from, which is what makes editing and reference-guided generation possible alongside generation from a text prompt.

Those capabilities point to a fairly broad set of practical uses. You can generate images directly from text prompts and control the shape, quality and background of the result, including transparent backgrounds for assets that need to be composited later. Because the model accepts image and file inputs, you can also use it for image editing or transformation tasks where the source image is passed in as a reference, and for workflows that mix text reasoning with image output, such as producing a written explanation alongside a generated or edited image. The reasoning and structured output parameters suggest it can also be used in pipelines where the text portion needs to be parsed programmatically.

You call the model through api14's OpenAI-compatible API using the model id "openai/gpt-5-image". The supported request parameters are include_reasoning, logprobs, max_tokens, reasoning, response_format, seed, structured_outputs, temperature, top_logprobs, top_p and verbosity. The image-specific controls (aspect ratio, quality, background, n, input references and output compression) are passed as part of the request as well. Because the interface is OpenAI-compatible, existing OpenAI SDK clients and tooling generally work with only the base URL and model id changed.

There is no information here about rate limits, latency or availability, so treat those as unknowns to test against your own workload. If your task is purely text and never touches images, a text-only model in the GPT-5 family may be a simpler fit. If you need image output with strong instruction following and reference images, this model is the one to evaluate first.