What happened
OpenAI’s model documentation lists GPT-Image-2.6 Sunburst under the identifier gpt-image-2.6-sunburst and describes it as a model for image generation and editing. The page says it accepts text and image input and produces image output. It positions the model for detailed compositions and image-editing workflows. A separate guide linked from the page covers the complexity setting, markup-guided edits, SVG output and animated GIF output. The supplied material also describes GPT-Image-2.6 Flare as released alongside Sunburst, but provides no model page or technical specifications for Flare here.
The documentation enumerates token prices for Sunburst. Text input costs $5 per million tokens, or $1.25 per million cached input tokens. Image input costs $8 per million tokens, or $2 per million cached input tokens. Image output costs $30 per million tokens. Those are the published unit prices; the source does not give a typical request cost, a per-image price, or a worked estimate for a particular resolution, prompt, or edit. Actual application cost therefore depends on usage that this excerpt does not quantify.
The model page lists both v1/images/generations and v1/images/edits among the available API endpoints, alongside other platform endpoints. It also lists Chat Completions and Responses as live endpoints. The source establishes these as documented interfaces, but does not specify which endpoint is required for each feature in the guide, nor does it provide implementation examples in the supplied excerpt. Builders should consult the linked image-generation guide for the actual parameter and request details rather than infer them from the model listing.
The page does not publish a performance rating or speed estimate: both fields are shown as not specified. It states that image generation and editing are supported, while audio and video are not supported. Rate limits vary with the organization’s usage tier, and the page directs customers to their account’s organization limits for applicable caps. The supplied evidence does not state a launch date, announce a change from an earlier version, or describe availability outside the documented API surface.
What the archive adds
The archive has covered Canva Craft’s preview developer workflow, but that is a separate provider and product, not prior coverage of this OpenAI release. This is a new event in image-generation APIs, while adding another example to the broader application-infrastructure thread. A Telegram post on September 9 described ChatGPT Images 2.5 and GPT-Image-2.5; the new material here is documentation for a 2.6 API model and its listed price and capabilities, not proof of a direct migration or upgrade path.
Trend signal
This supports the pattern that image generation is being exposed as application infrastructure: generative media becomes application infrastructure. A named model, API endpoints, token pricing and editing modes give application teams concrete integration surfaces to evaluate. The evidence is limited to a model page and linked guide; it does not establish real-world cost, latency, reliability or whether developers can build a durable product around these capabilities.
Why now
The practical timing is the documentation of a new numbered model and a set of generation and editing capabilities that builders can evaluate now. The supplied source does not explain why OpenAI is releasing the models at this time, state a launch date, or connect the release to an external event. No broader timing rationale can be established from the material.
Connections to other events
OpenAI’s September 9 ChatGPT Images 2.5 and GPT-Image-2.5 announcement, described in the archive’s earlier Telegram material, is a related product thread: that post concerned the 2.5 generation, while the current source documents the Sunburst 2.6 model for API use. Separately, the archive’s Canva Craft note covers another provider’s preview workflow for adding image generation to an app. These are comparisons of adjacent developments, not evidence that the products share technology, access terms or an integration path.
Why it matters
For teams building visual features, the release gives a documented model identifier, input and output modalities, endpoint names and per-token prices to use when scoping an integration. The listed editing endpoint and guide features suggest that the workflow need not be limited to producing a new raster image from a text prompt: builders can evaluate image input, guided edits, SVG output and animated GIF output against their product requirements.
The price figures help teams construct an initial budget model, but they are not enough to predict cost per finished asset. A team needs to measure token consumption for its own prompts, input images, outputs and caching behavior. The documentation’s lack of speed and performance figures also means production decisions should depend on hands-on tests, not assumptions based on the model name or supported formats. In short, the release expands the option set; it does not by itself establish that Sunburst is the right choice for a particular application. Later coverage: OpenAI released ChatGPT Images 2.6 with native GIF generation from text or photo prompts.
Limits and uncertainty
The supplied source does not provide a release date, benchmark results, quality comparisons, latency, reliability data, or representative costs per image or editing task. It does not explain the mechanics or limits of the complexity setting, markup-guided editing, SVG output, or GIF output beyond naming them in the linked guide. Pricing is expressed in tokens, not as a fixed per-image rate, and rate limits depend on the customer’s usage tier. The material names Flare but supplies no corresponding specifications, so claims about how it differs from Sunburst would be unsupported. No evidence here establishes a migration path from GPT-Image-2.5.
Skill to build now
Create a small evaluation set from the visual tasks your application actually needs: new compositions, edits to supplied images, and any required SVG or animated-GIF outputs. For each task, record the input, requested result, token usage, and whether the output meets your acceptance criteria. This lets the published token rates inform a real cost estimate and exposes quality or format gaps before you commit to a production workflow. The source supports testing these capabilities, but not assuming they meet any particular quality bar.
Reader takeaway
GPT-Image-2.6 Sunburst is documented as an API model for image generation and editing, with token-based pricing and additional capabilities surfaced in OpenAI’s guide. Treat the release as an option to test—not as evidence of superior quality, speed or economics. Measure it on representative tasks and check your organization’s limits before designing around it.