OpenAI GPT Image 2 for high-fidelity text-to-image generation and reference-based image editing from natural language instructions.
Added Apr 21, 2026
Approx. Price
$0.066 per image
Model Type
both
Settings
Generation controls available for this model.
Images Per Run
Up to 4
Output images
Input Images
Up to 4
Reference/edit images accepted • Route max 30 MB
Output Sizes
Aspect Ratio
Default
1:1
Custom Resolution
Number of Images
Default
1
Quality
Default
medium
Options (3)
Low, Medium, High
Quality of the generated image
Resolution
Default
1024x1024
Options (7)
1024x1024 (1:1 (Square)), 1024x768 (4:3 (Landscape)), 1024x1536 (2:3 (Portrait)), 1536x1024 (3:2 (Landscape)) +3 more
Benchmarks
Benchmarks
Human preference benchmarks sourced from Artificial Analysis.
Text to Image
#1 / 157
ELO
1178.0
Appearances
14,585
95% CI
-10/10
Image Editing
#2 / 74
ELO
1117.0
Appearances
11,046
95% CI
-10/10
Release Date 2026-04 · Matched as GPT Image 2 (high)
Artificial Analysis APIRelated image models
Compare GPT Image 2 with similar models from the same provider or model family.
GPT Image 1.5
gpt-image-1.5OpenAI's latest image model with better instruction following and adherence to prompts. Up to 4x faster than its predecessor with improved text rendering and iterative editing capabilities.
GPT 4o Image (old)
gpt-4o-imageOpenAI's GPT-4o image generation model. Supports both text to image and image-to-image generation.
GPT 4o Image
gpt-image-1OpenAI's GPT Image 1 model supports both text-to-image and multi-image editing. Can generate images from text or create new combined images from multiple inputs. Pricing includes input token costs.
DALL-E-3
dall-e-3OpenAI's first image model.
DALL-E-3 HD
dall-e-3-hdOpenAI's most well-known image model, now in HD quality.
MiniMax H3 Image Edit
wavespeed-ai/minimax-h3/image-editEdit images from up to nine references, preserving identity while changing scenes, outfits, or styles. Supports 1K and 2K output.