GPT Image 2.5

by openai

OpenAI’s current image generation and editing models, served through the Image API and the Responses API.

Models (OpenAI docs, 2026-10-02)

  • gpt-image-2.5-sunburst: for workflows where editing precision matters most.
  • gpt-image-2.5-flare: fast, high-quality everyday generation.
    Released 2026-09-08 (OpenAI API changelog, which also says the new xhigh and max quality settings arrived with it and token rates match GPT Image 2).

Capabilities

  • Endpoints: generations and edits (with masks); multi-turn editing in the Responses API via previous_response_id or image IDs.
  • Quality levels: low, medium, high, xhigh, max, auto.
  • Sizes: 1024x1024, 1536x1024, 1024x1536, or custom within 1:3 to 3:1 aspect ratios.
  • Streaming partial images (0-3); moderation parameter (auto or low).

Lifecycle of older image models

  • dall-e-2 and dall-e-3 shut down 2026-05-12.
  • gpt-image-1-mini, gpt-image-1.5 and chatgpt-image-latest are scheduled for shutdown 2026-12-01; migrate to gpt-image-2.5-sunburst or -flare.
  • gpt-image-2 and gpt-image-1 are named as replacements for DALL-E in the docs.

Billing

Token-based (text input, image input, image output; cost per image depends on quality and size). The Batch API is 50% off standard rates; cached-input rates apply only to images generated through the Responses API. Current prices: see the vendor pricing page https://developers.openai.com/api/docs/pricing

OpenAI - dall-e (discontinued predecessor) - nano-banana, midjourney, flux-dot-1, seedream-40, openai-sora

Sources (accessed 2026-10-02)