AI&Design
All archives
MARCH 2025 — NOVEMBER 2026 · KST

GPT Image

Track OpenAI’s GPT Image models from native image generation in GPT-4o to reference-led editing and animated or vector delivery. Compare the OpenAI, Runway, and Higgsfield interfaces on one timeline.

Posts archived
Attachments examined
Capabilities with evidence

What changed, when

Announcement dates in KST · Select a change or open its example.

Result example attached Announcement or contextual mediaBlank lane · no qualifying announcement in this source

Workflow threads

Read the dated updates and sources
  1. · GPT-4o generates and edits images natively

    Generate images in conversation, render instructions and text, and use uploaded visuals as references for further edits. OpenAI rolled the feature out in ChatGPT. GPT Image 1 API availability followed separately.

  2. · GPT Image 1 reaches the Image API

    Use OpenAI’s image model programmatically for generation and editing, with output settings including transparent backgrounds. OpenAI listed access requirements for developers and separately described future Responses API support.

  3. · Images 1.5 improves precise editing

    New targeted edits preserve more source details, render requested text more reliably, and generate faster than GPT Image 1 in OpenAI’s comparison. The release covers both ChatGPT Images and GPT Image 1.5 in the API.

  4. · Images 2.0 adds thinking and stronger composition

    OpenAI improves realism and layout, and introduces a ChatGPT thinking workflow that can reason or use web search before generating. Reasoning options refer to the ChatGPT release. API and Codex access use the GPT Image 2 model.

  5. · Higgsfield adds GPT Image 2.0

    Use GPT Image 2.0 in Higgsfield’s image tool for photorealistic visuals, packaging, and mockups at up to 4K. The notice describes product access via the Image model picker.

  6. · Runway Dev brings GPT Image 2 to its API

    Use GPT Image 2 for text-to-image or image-led generation via Runway’s text-to-image route. Runway’s dated changelog identifies gpt_image_2. Input settings depend on the product API.

  7. · Images 2.5 adds sketch input and two API variants

    Use sketches as references and place targeted comments on images in ChatGPT. The API adds Flare for faster iterations and Sunburst for edits that require tighter control. Sketch, templates, and on-image comments are ChatGPT tools. API capabilities differ by model.

  8. · Runway Dev lists Flare and Sunburst 2.5

    Runway exposes both OpenAI 2.5 variants for text and reference-led image generation. Each variant uses a separate model ID and Runway’s own generation settings.

  9. · Higgsfield offers Flare and Sunburst in Image

    Higgsfield presents Flare for everyday generation and Sunburst for precise edits and reference preservation. The product notice attributes quality and speed comparisons to OpenAI.

  10. · GPT Image 1 exits the OpenAI API

    OpenAI shuts down GPT Image 1 for API calls. The retirement applies to the original API model identifier.

  11. · Images 2.6 adds GIFs, markup, and SVG

    Animate images as GIFs, guide local edits with markup, and request vector output through the API. Flare and Sunburst remain separate model choices. GIF and SVG output use their own settings. Compare each downstream provider’s supported formats.

  12. · Higgsfield adds PNG, GIF, and SVG output

    Choose Flare or Sunburst and output PNG, GIF, or SVG, with transparent backgrounds available. Maximum GIF frame count depends on the output size. Model settings vary by output type.

  13. · Runway adds 2.6 PNG workflows

    Both variants offer PNG output including transparent backgrounds, up to 16 reference images, and configurable quality and reasoning effort. Runway’s November entries specify PNG for these routes. Up to ten outputs can be requested.

Documentation and availability notes
  • OpenAI Image API · · The documentation attributes each result to a version and distinguishes generated outputs from original image references.
  • OpenAI Image API · · The model matrix distinguishes supported outputs and options, including recent 2.6 settings, across OpenAI’s interfaces.
  • Runway Dev API · · Runway identifies separate endpoint model IDs and release dates for the GPT Image 2.5 and 2.6 Flare and Sunburst variants.
Coverage through November 20, 2026. Exact publisher timestamps are shown in KST; date-only notices retain the publisher's day.Back to capability timeline