OpenAI logo

GPT-Image-2

openai/gpt-image-2

Image model for prompt-driven generation, editing, and visual design workflows

Input modalitiesT
Output modalities
Source-linkedgpt-imageClosed weightsReleased 2026-04-21
Report data
Context windowNot documented
Maximum outputNot documented
Provider offers5
Knowledge cutoffUndisclosed
Capabilities

What this model supports

ReasoningNo
Tool callingNo
Structured outputNo
AttachmentsYes
Temperature controlNo
Open weightsNo
Creator
OpenAI
Family
gpt-image
Modalities
Image, Text
License
Not documented
Release date
2026-04-21
Model ID
openai/gpt-image-2
Market availability

Provider pricing

Report price →
Providers offering GPT-Image-2
ProviderProvider model IDContextInput / 1MOutput / 1MCached / 1MFeaturesAll pricing
Kenari logoKenarigpt-image-2272K$0 if offer.cost.input is defined else '—'}}$0 if offer.cost.output is defined else '—'}}$0 if offer.cost.cache_read is defined else '—'}}· · ·
View
{
  "input": 0,
  "output": 0
}
FastRouter logoFastRouteropenai/gpt-image-2128K$0 if offer.cost.input is defined else '—'}}$0 if offer.cost.output is defined else '—'}}$0 if offer.cost.cache_read is defined else '—'}}· · ·
View
{}
Vercel AI Gateway logoVercel AI Gatewayopenai/gpt-image-2Not documented$5 if offer.cost.input is defined else '—'}}$30 if offer.cost.output is defined else '—'}}$1.25 if offer.cost.cache_read is defined else '—'}}· · ·
View
{
  "cache_read": 1.25,
  "input": 5,
  "output": 30
}
Azure logoAzuregpt-image-2Not documented$5 if offer.cost.input is defined else '—'}}$30 if offer.cost.output is defined else '—'}}$1.25 if offer.cost.cache_read is defined else '—'}}· · ·
View
{
  "cache_read": 1.25,
  "input": 5,
  "output": 30
}
Poe logoPoeopenai/gpt-image-2Not documented$5.051 if offer.cost.input is defined else '—'}}$32.323 if offer.cost.output is defined else '—'}}$1.263 if offer.cost.cache_read is defined else '—'}}· · ·
View
{
  "cache_read": 1.2626,
  "input": 5.0505,
  "output": 32.3232
}
Evidence

Published benchmark results

Insufficient benchmark coverage. ModelBench does not infer a score from price, context size, or model name.