GPT Image 2 (ChatGPT Images 2.0)

2026-04-21 · Publisher : OpenAI · Domain : Artificial Intelligence

OpenAI launches ChatGPT Images 2.0 and the gpt-image-2 model: OpenAI's first image model with reasoning, outputs up to 2K and better rendering of non-Latin text.

What changes

  • Thinking mode: the model plans the image, can search the web and checks its output.
  • Up to eight consistent images from a single request, several formats, resolution up to 2K.
  • Better multilingual understanding and rendering of Japanese, Korean, Chinese, Hindi and Bengali.
  • Available in ChatGPT, Codex and the API; Thinking features reserved for Plus, Pro and Business.
  • Replaces dall-e-2 and dall-e-3, removed from the API on May 12, 2026.

### Specifications - Text and image input, image output. Flexible sizes, high-fidelity image inputs. - API: Images (generations, edits) and the Responses image generation tool; Batch with a 50% discount. - Transparent background in preview since August 20, 2026. - Identifiers: gpt-image-2, gpt-image-2-2026-04-21. ### Pricing | Modality | Input | Cached input | Output | | --- | --- | --- | --- | | Image | $8.00 | $2.00 | $30.00 | | Text | $5.00 | $1.25 | n/a | Price per million tokens (USD), current standard pricing; Batch: $4 and $15 for image, $2.50 for text. ### Benchmarks No numerical score published by the vendor in the announcement. ### Key takeaways Dense, multilingual text rendering and prior verification make the model usable for production visuals (infographics, mockups). Token-based billing makes cost per image harder to estimate.

Links

Tags : Image, OpenAI, GPT Image, Génération d'images, Raisonnement