GPT Image 2

Released
2026-04-21
Last refreshed
2026-10-07
Status
Researched 187d ago
ProprietaryCommercial use: restrictedMultimodal
OpenAI releases · 49 in the last 12 months · this family litChangelog →
Specifications
Family
GPT Image
Released
2026-04-21
Architecture
Autoregressive Image
Specialization
image-generation
Openness
Proprietary
License
ProprietaryRestricted
Weights
Not released
Code
Unknown
Training
Pretrained
Created by

Cutting-edge research and development.

San Francisco, California, United States
Founded 2015
Website
Pricing
Output / 1M
$30.00
Input / 1M
$5.00

Cheapest of 3 routes · Vercel AI Gateway · cache read $1.25

About

OpenAI's image generation model succeeding DALL-E 3, released April 21, 2026. Uses an autoregressive architecture with native reasoning — the model plans structure and composition before generating pixels. Supports up to 4K (4096×4096) resolution, achieves ~99% character-level text accuracy across Latin, CJK, Hindi, and Bengali scripts, and generates images ~2× faster than DALL-E 3. Powered by the GPT-5.4 backbone. API: $8/M image input tokens, $2/M cached, $30/M image output tokens. ChatGPT GA April 22, 2026; API access May 2026.

Providers(3)

Compare all 3

Compare API pricing across 3 tracked provider routes for input and output tokens, batch, and cached reads when available.

ProviderInput / 1MOutput / 1MCacheRoute
Vercel AI Gateway$5.00$30.00read $1.25
Serverless
OpenAI API$8.00$30.00read $2.00
Serverless
fal API---
ServerlessPartial

Available via routers & gateways(15)

Capabilities

VisionMultimodalReasoning

Benchmark peer barsfor Coding

No task-mapped benchmark peers are available for this model yet.

Benchmark scores(1)

Scores are benchmark-specific and are direction-aware: the same numeric gap can mean very different outcomes across suites. Use the leaderboard context and this model's provider route to decide whether the winning margin is meaningful for your workload.
BenchmarkScoreVersionEvaluationSource
1,381text_to_image overall 2026-10-26Observed 2026-10-27
Configuration: GPT-Image-2 (Medium)
Harness: LMArena blind pairwise human votes, Bradley-Terry
Evaluator: LMArena
Confidence: confirmed
Notes: LMArena text to image leaderboard, overall category, published 2026-10-26 (HF lmarena-ai/leaderboard-dataset @a467e6b9, latest split). Board entry 'GPT-Image-2 (Medium)': rank 3 of 82, Arena Score 1381.3 (95% CI 1377.4-1385.1), 103385 votes. Board entry carries the Medium quality qualifier. Source: LMArena (CC BY 4.0). Retrieved 2026-10-27.

Migration checks

No linked migration route is available for this model yet.

Rankings & picks(1)