What is GPT-Image-2.5 Sunburst?
GPT-Image-2.5 Sunburst is an image generation and editing model provided by OpenAI. It is designed for applications that need more than a quick text-to-image result: examples include detailed creative production, high-fidelity transformations, advertising artwork, diagrams, infographics, and edits where the subject's identity or composition needs to remain consistent.
The model accepts both text and image inputs. A text prompt can describe a new image, while an input image can provide material for an edit or transformation. The output is an image rather than a conventional text response, so Sunburst is best understood as a visual production model rather than a general-purpose conversational model.
OpenAI positions Sunburst as the most capable model in its GPT Image 2.5 lineup for precision-oriented work. That is a provider positioning claim, not an independent benchmark result. The practical implication is that users should consider Sunburst when quality, detail retention, and adherence to a visual brief matter more than the shortest possible generation time.
Where Sunburst fits in OpenAI's lineup
Sunburst belongs to the GPT-Image-2.5 family and is aimed at the higher-precision end of that family. The supplied OpenAI documentation describes GPT-Image-2.5 Flare as the faster alternative for everyday image generation. This creates a straightforward choice:
- GPT-Image-2.5 Sunburst: choose it for detailed editing, visual accuracy, complex layouts, and work where avoiding repeated attempts is valuable.
- GPT-Image-2.5 Flare: consider it when rapid iteration and lower generation latency are more important than maximum precision.
Sunburst is not presented as a general language model, an audio model, or a video model. Its specialization is image creation and image editing.
Core capabilities and quality controls
Sunburst supports two principal workflows: generating an image from a text description and editing an existing image using text instructions. Image editing can be useful for changing a scene, adjusting visual elements, transforming a style, or refining a composition without starting from an entirely blank canvas.
The model supports the following quality settings:
lowmediumhighxhighmaxauto
These settings provide a quality-versus-cost-and-speed control, although the supplied documentation does not provide a fixed generation-time or token-consumption figure for each setting. In general, applications should treat higher-quality settings as choices for important final outputs rather than automatically using the maximum setting for every draft.
Sunburst also supports flexible image-size controls subject to OpenAI's documented constraints. The exact available dimensions and restrictions should be checked against the current API documentation before implementation because the supplied research does not specify a universal maximum size.
Supported inputs and outputs
Sunburst supports text input and image input, and it produces image output. It does not support audio or video input or output according to the supplied model documentation.
| Modality or feature | Support |
|---|---|
| Text input | Supported |
| Image input | Supported |
| Image output | Supported |
| Audio input or output | Not supported |
| Video input or output | Not supported |
| Text as the primary output | Not supported |
The absence of a conventional text response matters when designing an application. Sunburst can create the visual result, but it should not be selected as the sole model for a workflow that needs extensive written reasoning, natural-language answers, or text-based agent behavior.
API access and integration
The canonical API model ID is gpt-image-2.5-sunburst. Developers can select it directly in the Images API for standalone image generation or editing requests. It can also be used through the image-generation tool in the Responses API, which supports multi-turn workflows and iterative image editing.
The Images API is the more direct option when an application mainly needs an image request and an image result. The Responses API image-generation tool is more appropriate when image creation is one step in a broader multi-turn interaction. The model documentation also notes that organization verification may be required before GPT Image models can be used through the API.
Sunburst does not support streaming, function calling, structured outputs, or fine-tuning. These restrictions make it unsuitable as the visual component of an agent that depends on native function calls or machine-readable structured responses from the image model itself. An application can still combine image generation with other software components, but those orchestration features would need to be handled outside Sunburst.
GPT-Image-2.5 Sunburst pricing
OpenAI prices Sunburst by tokens rather than by a single fixed amount per generated image. The listed rates are:
| Token category | Price per 1 million tokens |
|---|---|
| Text input | $5.00 |
| Cached text input | $1.25 |
| Image input | $8.00 |
| Cached image input | $2.00 |
| Image output | $30.00 |
The final cost depends on the prompt, any supplied image, output resolution, selected quality, and the number of image tokens used. Consequently, the $30.00 image-output rate should not be interpreted as a flat $30 charge for one image; it is a rate per million image output tokens.
OpenAI's supplied documentation states that its GPT Image 2 calculator does not estimate token consumption for GPT Image 2.5. Teams should therefore measure usage with representative prompts and image sizes before setting production budgets. Repeated edits, high-quality settings, and large or detailed inputs can all affect total consumption.
Context and output limits
No context-window length or maximum output-token limit is provided in the supplied model research. Those fields should therefore be treated as unspecified rather than assumed to be unlimited or equivalent to the limits of OpenAI's text models.
The relevant output constraint for Sunburst is visual rather than a conventional text-token response. OpenAI documents quality and image-size controls, but the supplied material does not state a single maximum image dimension or a universal maximum number of generated image tokens for every request.
Reasoning, coding, and tool capabilities
Sunburst is not documented as a reasoning model with a published reasoning score, and no reasoning benchmark is supplied. It is also not a coding model and has no published coding score in the supplied research. A text model or an application layer may be needed to interpret requirements, write code, validate results, or explain an image-generation workflow.
The model supports image generation through OpenAI's Images API and image-generation tool, but it does not support function calling or structured outputs. It also does not support streaming. These are verified capability limitations from the supplied model data, not judgments about whether an external application can build equivalent orchestration around the model.
Main strengths and limitations
Strengths
- Supports both text-to-image generation and image-to-image editing.
- Targets detailed visual work and high-fidelity image transformations.
- Offers quality controls ranging from low through max, plus auto.
- Can be used through both the Images API and the Responses API image-generation tool.
- Fits workflows involving visual layouts, infographics, advertising artwork, and composition-sensitive edits.
- Can be a better choice than a speed-optimized image model when fewer failed or inadequate attempts are more valuable than minimum latency.
Limitations
- Generation may take longer than GPT-Image-2.5 Flare, according to OpenAI's positioning.
- Token-based pricing can make costs vary considerably between requests.
- It does not support audio or video modalities.
- It does not provide conventional text output as its primary result.
- Streaming, function calling, structured outputs, and fine-tuning are not supported.
- No context-window or maximum-output-token specification is provided in the supplied research.
- The GPT Image 2 calculator does not estimate token consumption for GPT Image 2.5.
When to choose GPT-Image-2.5 Sunburst
Choose Sunburst when the image itself is the important deliverable and the task benefits from precision. It is a strong fit for a campaign image that must follow a detailed brief, an infographic with a carefully planned composition, a product or character edit that should preserve important visual details, or a creative workflow where each additional retry costs time and money.
Sunburst is also appropriate when an application needs both creation and editing rather than only one of those functions. The ability to provide an image input makes it suitable for iterative workflows in which a user starts with an existing asset and progressively changes it.
Choose the faster GPT-Image-2.5 Flare option instead when the main requirement is quick drafting, high-volume experimentation, or short feedback cycles and the task does not demand the highest available precision within the GPT Image 2.5 lineup. Choose a separate text or multimodal language model when the primary requirement is written reasoning, code generation, tool calling, or structured machine-readable responses.
Bottom line
GPT-Image-2.5 Sunburst is a specialized OpenAI image model for detailed generation and editing. Its defining trade-off is precision over speed: it is intended for high-fidelity visual work, while GPT-Image-2.5 Flare is positioned for faster everyday generation. Before adopting Sunburst, developers should account for token-based pricing, variable image consumption, the absence of streaming and function calling, and the lack of audio, video, structured-output, and fine-tuning support. For applications where visual accuracy and editing quality outweigh rapid, inexpensive drafts, Sunburst is the more appropriate choice.