Introducing ChatGPT Images 2.5
OpenAI released ChatGPT Images 2.5, improving multi-turn instruction following, response speed, and subject preservation from reference photos Two new API model variants introduced: gpt-image-2.5-sunburst (precision-focused) and gpt-image-2.5-flare (speed-focused) OpenAI reports over 3 billion images generated across ChatGPT Images and the GPT-Image API models The update enables reference image input, allowing users to edit existing images with natural language prompts while preserving key subje
Analysis
TL;DR
- OpenAI released ChatGPT Images 2.5, improving multi-turn instruction following, response speed, and subject preservation from reference photos
- Two new API model variants introduced: gpt-image-2.5-sunburst (precision-focused) and gpt-image-2.5-flare (speed-focused)
- OpenAI reports over 3 billion images generated across ChatGPT Images and the GPT-Image API models
- The update enables reference image input, allowing users to edit existing images with natural language prompts while preserving key subjects
Why It Matters
OpenAI's continued iteration on image generation models signals intensifying competition in the multimodal AI space, with practical improvements in instruction following and reference image preservation making the technology more viable for professional workflows. The dual-model strategy (Sunburst vs. Flare) reflects a maturing approach to balancing quality and speed, giving developers clearer options for different use cases.
Technical Details
- Two new model IDs available via API:
gpt-image-2.5-sunburstoptimized for editing precision andgpt-image-2.5-flareoptimized for fast, high-quality everyday generation - Improved multi-turn instruction-following capability, allowing more coherent and consistent image edits across iterative prompts
- Enhanced reference photo subject preservation, enabling users to pass one or more reference images (via
-iflag) and receive edited outputs that maintain key visual elements - CLI tooling already adapting to support the new reference image input functionality, as demonstrated by community-built wrappers
Industry Insight
- The Sunburst/Flare split suggests OpenAI is targeting both professional/editing workflows and casual/high-volume use cases, signaling that image generation is moving from novelty to production tooling
- Reference image preservation improvements lower the barrier for enterprise adoption, particularly in design, marketing, and content creation pipelines where brand consistency matters
- The 3 billion image milestone indicates rapid scaling; competitors should expect continued pressure on quality and speed, making API differentiation increasingly important
Disclaimer: The above content is generated by AI and is for reference only.