GPT Image 2.6 API
OpenAI’s GPT Image 2.6 family pairs Sunburst for detailed visual work with Flare for fast iteration, supporting image generation, editing, SVG graphics and animated GIFs.
GPT Image 2.6 Flare API
openai/gpt-image-2.6-flare
GPT Image 2.6 Flare is designed for fast image generation and creative iteration, with reference-image editing, complexity control, markup-guided edits, SVG graphics and animated GIF output.
GPT Image 2.6 API Background
Overview
GPT Image 2.6 is OpenAI’s image-generation and editing family, with Sunburst for detailed visual work and Flare for fast iteration. Text instructions and reference images guide the visual result. The family covers still-image production, targeted revisions, vector graphics and short animated loops.
Development History
The 2.6 release introduces complexity control, markup-guided editing, SVG export and animated GIF output. This family page brings the Sunburst and Flare variants together so creators can select a model by the needs of each stage of a project.
Key Innovations
- Sunburst emphasizes detailed compositions and visual refinement; Flare emphasizes rapid creative iteration.
- Markup lets an editing request indicate where a requested change belongs.
- SVG and GIF outputs extend the workflow beyond a single raster image.
GPT Image 2.6 API Technical Specifications
Architecture
Both variants generate and edit images from text and image inputs. Use the exact variant identifier in requests: gpt-image-2.6-sunburst or gpt-image-2.6-flare. GPT Image 2.6 is the family name, not a replacement for either API identifier.
Capabilities
- Text-to-image generation and reference-image editing through /v1/images/generations and /v1/images/edits.
- complexity supports low, medium and high, with high as the default. It controls planning separately from quality.
- Animated GIF output supports 2–16 frames, with 9 by default. Frame duration defaults to 200 milliseconds.
- SVG output produces static vector graphics suitable for icons and illustrations.
- PNG, JPEG and WebP support conventional raster-image delivery.
Limitations
- SVG generation uses n=1 and does not support image editing, streaming or compression options.
- GIF output does not support partial-image output or custom output_compression.
- The models do not accept audio or video inputs.
- Text placement and repeated character details still benefit from explicit instructions and checking the generated result.
GPT Image 2.6 API Performance
Strengths
- Sunburst suits detail-oriented final visuals, product compositions and careful editing.
- Flare suits exploring multiple directions before choosing a composition to refine.
- Reference images and focused edit instructions help maintain a coherent visual brief across revisions.
Real-world Effectiveness
For a product campaign, start by exploring background, lighting and composition, then choose one direction for refinement. For an illustrated character, establish the silhouette, colors and accessories before requesting alternate poses. For a poster, specify the intended hierarchy of title, image and supporting text. Keep these choices explicit in the brief so successive revisions have a stable target. Compare outputs at the same quality, size and prompt when choosing between variants; equal token rates do not imply equal token consumption per result.
GPT Image 2.6 API When to Use
Scenarios
- Campaign concepts: compare several visual directions, select a composition, and refine the chosen image for a consistent campaign.
- Product imagery: work from a reference product image while varying setting, composition or visual treatment.
- Poster and editorial illustration: describe the visual hierarchy, intended mood and placement of key elements.
- Icons and identity exploration: request SVG for vector shapes and inspect the resulting graphic in the target layout.
- Character animation: describe a small action and the intended loop for a short GIF, such as a wave or reaction.
Best Practices
- Choose Sunburst for detailed work or Flare for rapid iteration. Keep the exact lowercase model code in integrations.
- Separate the scene description from revision instructions. For a focused edit, state the target object and the desired change clearly.
- Choose output format by deliverable: raster for ordinary imagery, SVG for vectors, and GIF for short loops.
- Use a fixed brief when comparing variants. Change one creative variable at a time when diagnosing an unwanted result.
- Official token rates for both variants are $5 per million text input tokens, $1.25 cached text input, $8 image input, $2 cached image input and $30 image output. These are upstream reference rates, not a fixed per-image price.
- SVG billing includes the generated reference image plus authoring-model tokens. Image cost varies with the actual token usage.
- Official references: https://developers.openai.com/api/docs/models/gpt-image-2.6-sunburst ; https://developers.openai.com/api/docs/models/gpt-image-2.6-flare ; https://developers.openai.com/api/docs/guides/image-generation .