GPT Image 2.6 Sunburst is available through EvoLink's unified API for image generation and precise editing. Use reference images and masks to create product visuals, replace backgrounds, or refine selected details. API model ID: gpt-image-2.6-sunburst.
Test GPT Image 2.6 Sunburst output quality, estimate the token cost of one request, and generate a sample before you formally integrate.
Create image task
Fixed to gpt-image-2.6-sunburstImage generation and editing. Final charges follow actual token usage.
Reference images add input tokens to the request.
A still image in PNG, JPEG or WebP.
Pixel budget for the chosen ratio: 1K about 1.05 MP, 2K about 4.19 MP, 4K is 8.29 MP.
Rendering effort across five tiers, low to max. It drives the output token count, so it is the single biggest cost lever.
How much the model plans before rendering. Lower suits focused edits; higher helps with dense text and complex layouts. No extra charge; high is the default.
Transparent output includes an alpha channel. This preview feature may produce inconsistent results.
Results are kept for 24 hours — download anything you want to keep.
Start from the cost of a single sample, pick a testing budget, then price your own case in the calculator. The raw token rates are at the bottom.
First Test Cost
GPT Image 2.6 Sunburst - 1K - 1:1 - Low - 1 imageOne Low-quality image, no reference images
About 386 tests with $10 in credits.
For this estimate, use 1K, 1:1, Low quality and 1 image with the prompt “Cat”. Longer prompts and reference images add input cost; final charges follow actual usage.
GPT Image 2.6 Sunburst testing budget guide
Pick an amount based on how many tests you expect.Good for a first validation.
Good for prompt iteration.
Good before production integration.
Final charges are based on actual token usage.
Rates are per 1K tokens. The calculator estimates a full request; final charges follow actual usage.
gpt-image-2.6-sunburstNo fixed per-image price: cost depends on the prompt, size, quality, reference images and image count.
| Item | Rule | Rate | Billing |
|---|---|---|---|
| Image output tokens | Output token usage depends on image size and quality. | $0.030/1K tokens 2.04 cr/1K tokens | Output tokens |
| Image input tokens | Reference images and masks add image input tokens. | $0.0080/1K tokens 0.544 cr/1K tokens | Input tokens |
| Image cached input tokens | Charged when upstream usage reports cached image tokens. | $0.0020/1K tokens 0.136 cr/1K tokens | Input tokens |
| Text input tokens | Text tokens used by the prompt. | $0.0050/1K tokens 0.34 cr/1K tokens | Input tokens |
| Text cached input tokens | Charged when upstream usage reports cached text tokens. | $0.0013/1K tokens 0.085 cr/1K tokens | Input tokens |
Final charges are based on actual token usage.
GIF output uses the same token rates: all frames are tiled into one sheet and billed as one image of that size. The calculator shows the sheet and the price for each frame-count step.
Use the GPT Image 2.6 Sunburst API through EvoLink to generate detailed visuals or revise product and campaign images. For edits, send reference images with a written instruction and an optional mask or markup layer. Each revision is a separate asynchronous request.
Best fit: Precise reference edits, multi-turn revisions, campaign and product imagery · Cost basis: Prompt, size, quality, and reference-image input · Output boundary: Custom sizes up to a 3840 px edge; large outputs are experimental
gpt-image-2.6-sunburstPOST https://api.evolink.ai/v1/images/generations with model: gpt-image-2.6-sunburst and Authorization: Bearer $EVOLINK_API_KEY. The response contains id, not a finished image. Query GET /v1/tasks/{task_id} with that id.
These are EvoLink JSON request fields. Set quality and complexity separately. GIF options belong inside model_params; OpenAI SDK or other gateway examples may use a different field structure. Validate the combinations in the format table below.
| Parameter | Type | What it does |
|---|---|---|
| model | string · required | Use gpt-image-2.6-sunburst for this route. The display name and page slug are not the request model ID. |
| prompt | string · required | The text instruction. Prompt tokens are metered as text input on the token route. |
| size | string · default auto | 'auto', one of 15 aspect ratios such as 1:1 or 16:9, or explicit WxH pixels in multiples of 16. |
| resolution | string · 1K / 2K / 4K | Pixel budget for ratio mode; ignored when size is auto or explicit pixels. |
| quality | string · low / medium / high / xhigh / max | Rendering tier; default medium. EvoLink accepts the five explicit tiers shown here, not auto. Output usage varies with quality and dimensions. |
| complexity | string · low / medium / high | Planning effort, separate from quality; default high. No separate complexity fee. Set it explicitly when comparing requests; latency and total usage can still vary. |
| background | string · opaque / transparent | opaque by default. Transparent raster output is restricted to PNG/WebP and remains in Preview. |
| output_format | string · png / jpeg / webp / gif | png by default; svg is not supported yet. Use the format table for allowed inputs and batch counts. |
| n | integer · 1–10 | Static requests accept 1–10; the actual output count is results.length and can be lower. GIF requires 1. |
| image_urls | array · up to 16 | Server-accessible reference URLs. Order matters for a mask or markup: the first image is the editing base. References contribute image input tokens. |
| mask_url | string · optional | URL of an alpha-channel PNG for local regeneration. Requires a reference image; mutually exclusive with markup_url. |
| markup_url | string · optional | URL of a transparent PNG annotation layer for the first reference. Keep it separate from the photo; mutually exclusive with mask_url. |
| model_params | object · optional | GIF: frame_count 2–16 (default 9), frame_duration_ms 20–2000 (default 200); send them only with output_format gif. The SVG fields are not supported yet. |
| callback_url | string · optional | HTTPS webhook that receives the task result, as an alternative to polling. |
Use the same EvoLink credentials for a server-side integration or an agent-assisted call. Keep this route’s model ID and request fields in the configuration.
Best for: product backends, batch jobs, automated workflows
Call EvoLink’s unified image API from your server and control the model ID, parameters, task queue, callbacks, and result storage.
Best for: creative and development tasks in Codex, Claude, and Gemini
Give the Agent your output goal, assets, and acceptance criteria. It can choose the route, assemble the request, track the task, and return the result without requiring you to hand-code every step.
Reference-image edit example. Replace the example image URL with a server-accessible product photo. Save response.id and query the task; read results only after completed. An optional mask_url or markup_url requires that reference input.
curl -X POST https://api.evolink.ai/v1/images/generations \
-H "Authorization: Bearer $EVOLINK_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "gpt-image-2.6-sunburst",
"prompt": "Replace only the background with a pale gray studio. Preserve the product shape, colors and label text.",
"size": "auto",
"quality": "medium",
"complexity": "low",
"output_format": "png",
"n": 1,
"image_urls": [
"https://example.com/product.png"
]
}'
# Save id from the create response, then query that task:
curl https://api.evolink.ai/v1/tasks/{task_id} \
-H "Authorization: Bearer $EVOLINK_API_KEY"
# pending / processing: keep querying the same task.
# completed: download the URLs in results within 24 hours.
# failed: inspect error.code and error.message before retrying.Check this route’s enums and nesting. Do not send GIF fields with PNG output, combine mask_url with markup_url, or request n > 1 for GIF. output_format svg is not supported yet.
Use URLs the server can download. A mask needs alpha transparency; a mask or markup must match the first reference dimensions after orientation. Supply the reference as well as the layer.
Check the EvoLink Bearer key and available balance. Keep the key on your server or in your local environment.
If an id was returned, query that task before creating another. pending or processing is not a failure. For failed, inspect error.code and error.message; correct the cause before retrying. Failed tasks are not charged.
Use a public HTTPS callback_url that returns 2xx promptly. Keep the task id so your application can query the final state even if callback delivery fails.
| Request rule | Static generation | Reference-image editing | Animated GIF | SVG (not supported yet) |
|---|---|---|---|---|
| Request inputs | prompt | prompt + image_urls | prompt; optional image_urls | — |
| output_format | png / jpeg / webp | png / jpeg / webp | gif | svg |
| Results per request | n = 1–10 | n = 1–10 | n = 1 | — |
| Editing and format controls | size / resolution / quality / complexity | mask_url or markup_url | model_params: frame_count, frame_duration_ms; no mask_url | — |
| Billable work | Text input + image output | Text/reference inputs + image output | Input + frame-sheet generation | — |
Request validation checks
For explicit WxH, both edges must be multiples of 16 and at most 3840 px, with 655,360–8,294,400 total pixels and an aspect ratio from 1:3 to 3:1. Above 3,686,400 pixels is experimental. resolution only applies to ratio mode.
Use size: auto for 512×512 frames or explicit WxH in multiples of 16. Ratio strings and resolution are not used. The selected frame count must fit the frame-sheet limits; consult the API reference for the per-frame budget. This EvoLink route accepts 2–16 frames.
Set background: transparent with PNG or WebP and inspect the alpha channel; JPEG cannot preserve transparency. Transparent output remains in Preview. GIF on this route requires opaque output.
Up to 16 JPEG/PNG/WebP references, each at most 50 MB, using server-accessible URLs. A mask or markup must match the first image after orientation and requires a PNG with transparency. Keep the markup separate; it and its first reference must each be at most 16,777,216 pixels.
The create response returns id and status. Use id in GET /v1/tasks/{task_id}; keep querying while pending or processing. On failed, read error.code and error.message. A client timeout alone does not justify resubmitting a task that already has an id.
On completed, results is an array of file URLs. Save every returned file within 24 hours and use results.length as the delivered count. To continue an edit, pass an approved output URL as a new reference and retain a downloaded checkpoint.
The create response can show credits_reserved; final billing follows reported usage. Check the completed task record instead of treating the reservation as the final charge. The existing Pricing section supplies estimates and rates; references, output count and GIF frame sheets affect usage.
Use one EvoLink API key across your supported image and other model routes. Keep each route’s model ID and accepted fields in configuration so an application can reuse authentication without assuming identical request support.
Use one EvoLink balance and task history across your workflows. Current rates and request estimates stay in the existing Pricing module; per-task records help reconcile spending without maintaining separate provider billing integrations.
API model IDs and route pages for the current GPT Image variants and their 2.5 predecessors.
GPT Image 2.6 Flare
The fast GPT Image 2.6 variant: OpenAI's default for high-quality everyday generation on the same endpoint and rates.
View API
GPT Image 2.5 Sunburst
Previous-generation API reference for existing GPT Image 2.5 integrations.
View API
GPT Image 2.5 Flare
Previous-generation API reference for existing GPT Image 2.5 integrations.
View APIGPT Image 2.6 Flare
The other GPT Image 2.6 model: OpenAI's default for everyday generation and rapid iteration, same endpoint and token rates.
View API
Nano Banana 2
Google image model for high-quality generation and flexible creative workflows.
View API
Nano Banana Pro
Premium Nano Banana route for higher-fidelity images and production-grade output.
View API
Seedream 5.0 Pro
BytePlus image route with 1K/1.5K/2K tiers, multi-reference editing, and layer decomposition.
View API
Grok Imagine Image 2.0
xAI image route where the number of input images picks the mode, with 1K/2K output and up to 10 images per request.
View APIHistorical 2.5 task-selection guide. It does not establish 2.6 results or current route support.
The fast GPT Image 2.6 variant: OpenAI's default for high-quality everyday generation on the same endpoint and rates.
Previous-generation Sunburst API reference for existing integrations.
What OpenAI released on September 8, 2026 — the generation GPT Image 2.6 builds on.
EvoLink API calls use your EvoLink API key and balance. A ChatGPT subscription does not provide an EvoLink API allowance. Any trial credits depend on the benefits shown in your EvoLink account.
The documented static-image range is 1–10, but upstream support for more than one output is unreliable. Count results after completion and use the returned usage for billing. For a reliable batch workflow, submit and track separate n: 1 tasks; GIF always requires n: 1.
The documented EvoLink JSON request uses server-accessible image_urls, mask_url and markup_url. Upload a local file first and use its downloadable URL. A local path or an OpenAI file ID is not a replacement for a URL in these fields.
A mask, markup or preservation instruction guides the edit but does not guarantee pixel-identical output. Check the product, labels and boundaries. Composite the original product over the generated background when unchanged source pixels are mandatory.
Each revision is a new request using the reference you provide. Carry forward the preservation requirements, inspect the output and keep approved files. If a revision drifts, resume from an approved checkpoint or the original instead of using the degraded result.
No spelling or placement guarantee is established by these request controls. Quote required text and specify its position, then review the returned file at delivery size. Correct final text in a design tool when accuracy is mandatory.