ChatGPT Images 2.5 is a faster, sharper image-generation model in OpenAI's ChatGPT lineup, released September 8, 2026, and built for designers, marketers, and everyday creators who iterate on visuals inside chat. It is the consumer-facing upgrade to the ChatGPT Images line and the GPT-Image API models that developers call directly.
Core Features
- Up to 50% lower generation latency than Images 2.0, so multi-step edits finish in far less time.
- Stronger reference fidelity: it preserves a person's face or a product's shape while changing only the element you ask for.
- Multi-turn consistency, meaning an edit made three prompts ago survives later changes instead of drifting.
- New in-chat tools: Sketch lets you draw a rough layout as a visual reference, Templates cover posters and product shots, and inline Comments mark the exact spot to change.
- Two API models: GPT-Image-2.5 Flare for speed and GPT-Image-2.5 Sunburst for high-precision commercial work.
Use Cases
- Marketing teams producing campaign visuals and product mockups at scale.
- Designers who need reliable local edits without regenerating the whole image.
- Developers embedding image generation into apps through the OpenAI API.
- Creators exploring ideas with quick sketches rather than long text prompts.
Pricing
In ChatGPT, Images 2.5 is included on every tier, from free through ChatGPT Work and Codex, on desktop, mobile, and web. Via the API, OpenAI lists image input at $8.00 per 1M tokens ($2.00 cached), image output at $30.00 per 1M tokens, and text input at $5.00 per 1M tokens ($1.25 cached). Adobe Firefly and Midjourney remain paid alternatives without the same chat-native editing loop.
Our Take
Best for teams that already live in ChatGPT and want controlled, iterative editing rather than one-shot generation. The trade-off is that Sunburst's precision costs noticeably more latency, so high-volume jobs should default to Flare. Browse more Image Generation tools for comparisons.
