What is chatgpt-image-latest?
chatgpt-image-latest is an OpenAI API alias for the image-generation system previously used in ChatGPT. Rather than being a permanently versioned model family, the alias identifies a ChatGPT image snapshot that OpenAI made available for API use. OpenAI introduced it in connection with the updated ChatGPT Images experience on December 16, 2025.
The model’s primary purpose is visual creation and transformation. A developer can provide a written prompt to generate an image, or provide an existing image together with instructions for editing or transforming it. The model can return generated images and text, so it is not limited to a text-only response interface.
Although the alias is still accessible, OpenAI lists it as deprecated. The provider has announced a shutdown date of December 1, 2026 and recommends GPT-Image-2 as the migration target. For new API image-generation projects, OpenAI also positions GPT-Image-2.5 Sunburst and GPT-Image-2.5 Flare as current alternatives for different quality and speed requirements.
Where it fits in OpenAI’s lineup
chatgpt-image-latest belongs to OpenAI’s GPT Image family, but its role is transitional. The name indicates that it follows the image snapshot used in ChatGPT rather than representing a stable, independently dated model intended for long-term reproducibility.
That distinction matters for production systems. Existing applications may continue using the alias while it remains available, but behavior associated with a moving or ChatGPT-linked alias may be less suitable when an application needs repeatable results over time. Developers planning a new integration should normally select a currently supported canonical or dated image model instead of building around this deprecated name.
OpenAI’s documented migration direction is GPT-Image-2. Within the newer GPT Image 2.5 positioning described in the supplied research, Sunburst is aimed at high-fidelity generation and editing, while Flare is aimed at faster everyday image generation. Those models are relevant alternatives, but they do not change the fact that chatgpt-image-latest itself is the subject of this page: a legacy image alias that should be treated as temporary.
Inputs, outputs and supported modalities
chatgpt-image-latest supports text and image inputs. Text input covers prompts, instructions, and editing directions. Image input enables workflows such as changing an uploaded picture, creating variations, or using an existing visual as the basis for a new result.
Its output modalities are text and images. Image output is the central capability, while text can be returned through the model interface. The model does not provide native audio or video input or output, and it is not an embedding, speech, music, or action-output model.
- Text input: Supported for prompts and image instructions.
- Image input: Supported for editing and transformation workflows.
- Text output: Supported.
- Image output: Supported for generated and edited visuals.
- Audio and video: Not supported as native input or output modalities.
OpenAI documentation describes support for several image sizes, including 1024×1024, 1024×1536, and 1536×1024. This gives applications square, portrait-oriented, and landscape-oriented output options, although the cost varies with both quality and size.
What the model does well
The model is most useful when the task is primarily visual and can be expressed through natural-language instructions. Typical examples include generating a new illustration, creating marketing artwork, producing ecommerce or product imagery, and transforming an uploaded image into a different visual treatment.
- Text-to-image creation: Generate visuals from detailed descriptions of subjects, settings, composition, and style.
- Image editing: Modify an existing image according to an instruction rather than starting entirely from scratch.
- Creative prototyping: Explore visual concepts and variations before committing to a final design.
- Marketing and branding: Produce draft campaign assets, promotional concepts, and branded visual directions.
- Product visualization: Create or adapt product imagery for ecommerce and presentation workflows.
These are capability-based use cases supported by the model’s documented text-and-image interface. They should not be interpreted as a guarantee of a particular artistic style, visual accuracy, or production-ready result for every prompt.
Pricing and image-size trade-offs
OpenAI lists both token-based charges and separate per-image generation prices for chatgpt-image-latest. Token pricing applies to text and image content processed by the model, while per-image pricing depends on the requested image quality and size.
| Charge type | Published price |
|---|---|
| Text input | $5 per 1 million tokens |
| Cached text input | $1.25 per 1 million tokens |
| Text output | $10 per 1 million tokens |
| Image input | $8 per 1 million tokens |
| Cached image input | $2 per 1 million tokens |
| Image output | $32 per 1 million tokens |
For image generation, the documented per-image prices are:
| Quality | 1024×1024 | 1024×1536 or 1536×1024 |
|---|---|---|
| Low | $0.009 | $0.013 |
| Medium | $0.034 | $0.05 |
| High | $0.133 | $0.20 |
The practical trade-off is straightforward: low quality is the least expensive option for drafts and rapid experimentation, while high quality costs substantially more and is better reserved for outputs where additional visual fidelity justifies the expense. The larger portrait and landscape formats also cost more than the square format at the same quality level.
These prices are usage-based API charges, not monthly subscription plans. The supplied research does not specify a context-window limit, maximum output-token limit, or a separate fixed subscription price for this model.
API support and technical limitations
chatgpt-image-latest is available through OpenAI image-generation and image-edit endpoints. The supplied documentation also identifies support for the Responses API image-generation tool and the Batch API. Batch support can be relevant for workloads that process many independent image tasks rather than requiring every result interactively.
Several common language-model features are not supported. The model does not support streaming, function calling, structured outputs, or fine-tuning. This limits its suitability for applications that require incremental response delivery, tool orchestration, machine-validated JSON, or custom model training.
- Image generation and editing endpoints: Supported.
- Responses API image-generation tool: Supported.
- Batch API: Supported.
- Streaming: Not supported.
- Function calling: Not supported.
- Structured outputs: Not supported.
- Fine-tuning: Not supported.
The absence of structured outputs is especially important for developers who need reliable machine-readable fields. Although the model can return text, it should not be selected as a general-purpose structured-data or workflow-control model.
Reasoning, coding and performance profile
chatgpt-image-latest is an image model rather than a general reasoning or coding model. The supplied model assessment gives it a reasoning score of 1 and a coding score of 1 on the catalog’s internal scale. These are editorial or catalog evaluations, not OpenAI-published benchmark results, and they should not be treated as formal performance measurements.
In practical terms, the model can interpret natural-language image instructions, but it is not intended for complex text reasoning, software development, code generation, or tool-driven automation. Its value comes from producing and modifying images, not from competing with general-purpose language models on analytical or programming tasks.
The same internal catalog rates its speed at 7 and cost at 5. These ratings are also editorial comparisons rather than provider specifications. They suggest a middle-ground profile within the catalog, but the actual experience and expense depend on image quality, resolution, input content, caching, and workload design.
When to choose chatgpt-image-latest
Choose chatgpt-image-latest primarily when an existing application already depends on the alias and needs continued access before the shutdown date. It can also be reasonable for a temporary migration phase, compatibility testing, or a short-lived workflow centered on image generation and editing.
Its strongest fit is an application that:
- Needs text-to-image generation or image editing.
- Can use the documented image sizes and quality tiers.
- Does not require streaming, function calling, structured outputs, or fine-tuning.
- Can tolerate the model’s deprecated status and planned removal.
- Benefits from image-generation and image-edit endpoints, Responses API image tools, or Batch API processing.
For a new production integration, another option is generally more appropriate. GPT-Image-2 is OpenAI’s stated migration target, while GPT-Image-2.5 Sunburst may be a better fit when high-fidelity generation and editing are the priority. GPT-Image-2.5 Flare may be more appropriate when faster everyday image generation matters more than maximum fidelity. These choices reflect the main trade-off: a current supported model reduces migration risk, while quality and speed requirements determine which newer option is most suitable.
Deprecation and migration planning
OpenAI has deprecated chatgpt-image-latest and scheduled API shutdown for December 1, 2026. Applications that use it should therefore separate image-generation logic from the rest of the product and test a replacement well before that date.
Migration testing should cover prompt behavior, image editing workflows, supported sizes, quality settings, output handling, cost, and any differences in how the replacement model responds to source images. If reproducibility is important, avoid assuming that the alias will remain behaviorally stable simply because its name remains available until the shutdown.
In summary, chatgpt-image-latest remains a capable interface for image generation and editing, but its deprecated status is the defining practical fact. It is best understood as a legacy bridge for existing ChatGPT image integrations, not as the foundation for a new long-lived API architecture.