GPT Image 2.6 — Flare and Sunburst
GPT Image 2.6 has two distinct model IDs. There is no gpt-image-2.6 model or alias.
| Model | Namespaced ID | Designed for |
|---|---|---|
gpt-image-2.6-flare | openai/gpt-image-2.6-flare | Image generation and editing for fast iteration |
gpt-image-2.6-sunburst | openai/gpt-image-2.6-sunburst | Image generation and editing for detailed visual work |
Use the IDs above, bare or with the openai/ prefix. Catalog listing does not establish serving availability; check your authenticated /models response.
Endpoints and request parameters
Use the synchronous Images API at https://api.quotaflow.ai/openai/v1/images/generations or /openai/v1/images/edits. These models do not support the Responses API: OpenAI's GPT Image 2.6 model pages, read November 4, 2026, list only the Images API endpoints.
quality:low,medium,high,xhigh,max, orauto. An omitted quality is not filled in. Unknown values return HTTP 400 withparam: "quality".size,n,mask,background,output_compressionandmoderationfollow the same rules as GPT Image 2.5.output_format:png,jpeg,webp,giforsvg. Only the two 2.6 models acceptgifandsvg. Other image models return HTTP 400 withparam: "output_format".complexity(low,mediumorhigh) is forwarded as sent.- SVG output (
output_format: "svg") is for generations only and requiressvg.model. That is the text model that writes the SVG, for example{"svg":{"model":"gpt-5.5","reasoning":{"effort":"low"}}}. Without it the request returns HTTP 400 withparam: "svg.model". Eachdataitem'sb64_jsonholds the SVG text (image/svg+xml). The response carriessvg_usagealongsideusage. No C2PA manifest is attached to SVG output. - Animated GIF output (
output_format: "gif") takesframe_count(2–16, default 9) andframe_duration_ms, both forwarded as sent.sizeis the size of one frame. All frames must fit on one sheet, which OpenAI sizes itself: 4 frames at512x512are served, while any frame count at1024x1024is refused. OpenAI's size refusal is returned as it was written. - SVG and GIF requests are served only by supplies that report token usage.
- OpenAI does not document
input_fidelityfor 2.6. Quotaflow applies the same handling as GPT Image 2 and 2.5 and does not forward it. - Not yet supported here:
markup-guided edits.
Omitting model still selects gpt-image-2.
curl https://api.quotaflow.ai/openai/v1/images/generations \
-H "Authorization: Bearer $QUOTAFLOW_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"gpt-image-2.6-flare","prompt":"A clean editorial illustration of an AI workflow.","size":"1024x1024","quality":"high","response_format":"b64_json"}'
Streaming uses the Images API SSE contract with n: 1; optional partial frames are capability-specific. Returned usage is the customer billing basis and may use validated reporting or supported estimates for the request path.
Standard token prices
Both models use the same standard rates. Prices are USD per 1 million tokens; Batch is not offered.
| Unit | Price |
|---|---|
| Text input | $5.00 |
| Cached text input | $1.25 |
| Image input | $8.00 |
| Cached image input | $2.00 |
| Image output | $30.00 |
An SVG answer is charged twice. Its usage is billed at the rates above. Its svg_usage is billed at the standard text rates of the model it names. For example, gpt-5.5 is $5.00 per 1M input tokens and $30.00 per 1M output tokens, and reasoning tokens are part of the output tokens. A contracted rate applies to both charges.
A temporary balance hold is separate from the final charge.
Official sources, read November 4, 2026: pricing, Flare, Sunburst, and image generation guide.