GPT Image 2.5 Flare

GPT Image 2.5 Flare is OpenAI's GPT Image 2.5 model for fast, high-quality everyday image generation. Its official API ID is gpt-image-2.5-flare. The model accepts text and image inputs and returns image output, supporting both prompt-based creation and reference-image editing. A dated gpt-image-2.5-flare-2026-09-08 snapshot is available when a workflow needs pinned behavior.

Flare can be selected directly through the Image API or used as the image_generation tool in the Responses API. It shares the family's main controls for custom dimensions, low through max quality, PNG/JPEG/WebP output, transparent backgrounds, compression, and streaming partial previews. Its positioning emphasizes rapid iteration and everyday creation; OpenAI recommends Sunburst when editing precision matters most.

Key Capabilities

• Fast Text-to-Image: Turns prompts into photography, illustrations, and design assets for everyday high-quality generation.
• Reference-Based Editing: Uses image and text inputs to adjust subjects, backgrounds, composition, style, and local elements.
• High-Frequency Iteration: Rapidly explores multiple composition, style, color, or channel-specific variations.
• Multi-Level Quality Control: Supports low, medium, high, xhigh, max, and auto.
• Custom Dimensions: Specifies precise width and height in addition to standard square, landscape, and portrait sizes.
• Transparent and Multi-Format Output: Returns PNG, JPEG, or WebP and supports transparent-background assets.
• Multi-Turn Image Workflows: Uses Responses API context to continue revising a visual direction over several turns.

Technical Strengths

Feature	Benefit
Speed-Oriented Family Tier	Serves everyday generation, previews, and frequently refreshed visual tasks as the faster 2.5 endpoint.
Unified Generation and Editing	Handles prompt creation and image-input revision without switching models between drafting and variation stages.
Image API and Responses API	Supports direct media requests as well as contextual tool calls and multi-turn applications.
Six Quality Settings	Explicitly spans quick drafts through higher-quality output or lets auto select for the request.
Flexible File and Canvas Controls	Supports custom dimensions, three formats, JPEG/WebP compression, and transparent or automatic backgrounds.
Streaming Generation Feedback	Returns up to three partial previews so interfaces can show progress before the final image is ready.

Frequently Asked Questions

How does GPT Image 2.5 Flare differ from GPT Image 2.5 Sunburst?

GPT Image 2.5 Flare targets fast, high-quality everyday generation, while Sunburst targets maximum capability and editing precision. Start with Flare for ideation, frequent content updates, and visual variants, and evaluate Sunburst first for complex local edits, multi-reference composition, or strict preservation requirements.

Does GPT Image 2.5 Flare support image editing and reference inputs?

Yes, GPT Image 2.5 Flare accepts text and image inputs for reference-based editing in addition to prompt-only generation. Use it for everyday background changes, style variants, and asset revisions, then compare with Sunburst when the edit region or referenced attributes must be retained more precisely.

Which custom dimensions does GPT Image 2.5 Flare support?

GPT Image 2.5 Flare accepts custom WIDTHxHEIGHT dimensions. Each edge must be a multiple of 16, the aspect ratio must remain between 1:3 and 3:1, neither edge can exceed 3840 pixels, and total pixels must stay between 655,360 and 8,294,400; dimensions above 2560×1440 remain experimental.

How can GPT Image 2.5 Flare balance quick previews with final output?

Start with a lower quality setting and a recommended standard size to validate the visual direction, optionally streaming up to three partial previews. After approval, submit a separate request at the desired final quality; partial images are progress feedback and should not be treated as finished deliverables.

How should GPT Image 2.5 Flare be connected after it appears on LinkModel?

First check the live LinkModel model page for the Flare platform ID, generation and edit entry points, reference-image fields, custom dimensions, quality settings, transparent-background controls, and response format. Reuse LinkModel's shared authentication and client setup only after support is confirmed rather than copying OpenAI-native IDs, URLs, or Responses tool structures.