한국어

Model Guide

GPT Image 2 — gpt-image-2

OpenAI launched GPT Image 2 on April 21, 2026. Powered by gpt-image-2 with Instant and Thinking modes, multilingual text, 2K output, and up to eight consistent images per prompt.

What is GPT Image 2?

GPT Image 2 is OpenAI's next-generation image system inside ChatGPT, Codex, and the API — released April 21, 2026. It replaces GPT Image 1.5 as the default and is powered by the gpt-image-2 model.

OpenAI describes it as a new era of image generation: stronger instruction following, production-usable outputs, readable multilingual text, and layout-aware reasoning for infographics, slides, product photography, maps, manga-style art, and more.

Instant mode vs Thinking mode

GPT Image 2 is not one single speed setting — pick the mode that matches your task:

  • Instant mode — fast generations with core model improvements. Available to all ChatGPT users.
  • Thinking mode — uses additional compute to plan composition, count objects, verify constraints, and optionally search the web before rendering. Available on Plus, Pro, Business, and Enterprise tiers.
  • Thinking mode can produce up to eight images from one prompt with consistent characters, objects, and style across the set.

Key capabilities (from OpenAI)

  • Readable multilingual text — Chinese, Japanese, Korean, Hindi, Bengali, and more with proper character structure
  • Up to 2K resolution with aspect ratios from ultra-wide 3:1 to ultra-tall 1:3
  • Native reasoning that plans layout before pixels are rendered (Thinking mode)
  • Web search during generation for current events and factual visual context (Thinking mode)
  • Character and object continuity across multi-image outputs
  • Infographics, slides, floor plans, image grids, and reference-image editing

Best use cases

  • Product photography — clean hero shots, lifestyle scenes, and packaging with readable labels
  • Infographics & diagrams — labeled charts, menus, and dense informational posters
  • Slide decks — presentation visuals with coherent layout and typography
  • Storyboards & ads — up to eight consistent frames for campaigns and social series
  • Multilingual marketing — posters and banners with accurate non-Latin copy
  • Spatial & technical — floor plans, maps, and architectural visualizations

How to try gpt-image-2 free online

ChatGPT Instant mode is free for all users. You can also try gpt-image-2 through free online generators — sign up with starter credits, no download required.

  1. GPT Image 2 생성기
  2. Create a free account
  3. Select GPT Image 2 as your model
  4. Enter your prompt and choose an aspect ratio
  5. Generate and download your image

Read OpenAI's announcement →

Try GPT Image 2 free

Start with Instant-mode quality — generate your first gpt-image-2 visual in under a minute.

무료로 생성 시작