Back to Models

GPT Image 1

openaiopenai/gpt-image-1

Previous generation image generation model.

GPT Image 1 (gpt-image-1) is OpenAI's first natively multimodal image generation model, built on a GPT-4-class decoder architecture with specialized visual token embeddings and cross-modal attention. Released in April 2025, it was the model that powered the initial ChatGPT image generation feature β€” which saw over 130 million users create 700+ million images in its first week alone. Unlike earlier diffusion-based approaches (DALLΒ·E 3), gpt-image-1 uses an autoregressive decoder that treats images as tokens, giving it world-knowledge grounding and the ability to follow intricate design instructions.

Through TheRouter, gpt-image-1 is accessible via the standard OpenAI-compatible Images API endpoint at api.therouter.ai/v1/images/generations. It supports text-to-image generation, image-to-image editing, inpainting with masks, and transparent-background output. The model is best suited for production pipelines needing reliable, high-fidelity visual content generation rather than experimental or open-source workflows.

Best for
  • β€’ Text-to-image generation β€” producing high-fidelity images from natural-language prompts across photorealistic, illustrative, anime, vector, 3D, and data-viz styles
  • β€’ Accurate text rendering inside images β€” posters, UI mockups, multi-line labels, and small-font typography that actually reads correctly
  • β€’ Image editing and inpainting β€” edit existing images with text prompts, apply masks for targeted changes, or transform images with new styles
  • β€’ Production-grade creative assets β€” social ads, hero images, e-commerce product renders, and brand-consistent marketing collateral
  • β€’ Enterprise-safe image generation β€” built-in C2PA provenance watermarks, tunable moderation, and no training on customer data
Reach for something else if
  • β€’ Ultra-low-latency single-shot drafts β€” DALLΒ·E 3 remains faster for quick 512px drafts when quality isn't critical
  • β€’ Fully open-source or offline deployment β€” unlike Stable Diffusion, gpt-image-1 is proprietary and only available via API
  • β€’ Experimental style exploration with broad community palettes β€” Midjourney v7 offers a wider variety of community-curated artistic styles
  • β€’ Real-time or streaming image generation β€” the model requires full prompt processing before returning results; no incremental output

How TheRouter serves this differently from the vendor

As the vendor operates it

OpenAI serves direct gpt-image-1 calls on /v1/images/generations and /v1/images/edits, and its model page also lists /v1/responses support for the model ID.

On TheRouter

TheRouter exposes image generation and image edit for openai/gpt-image-1 under https://api.therouter.ai/v1. Use the OpenAI SDK shape with TheRouter baseURL, API key, and routed model slug; do not assume chat/completions works for this image model.

Context Length
--
Max Output
--
Image Priceper 1M tokens
$0.0454/ image
Input Priceper 1M tokens
$5.40/ 1M tokens

Modalities

textimage→image

Capabilities

VisionImage GenerationImage Edit

Media Generation Capabilities

image_generation
sizes
  • 1024x1024
  • 1024x1536
  • 1536x1024
qualities
  • low
  • medium
  • high
max_outputs_per_request
4
output_formats
  • png
  • jpeg
  • webp
defaults
size
1024x1024
quality
medium
image_edit
sizes
  • 1024x1024
  • 1024x1536
  • 1536x1024
qualities
  • low
  • medium
  • high
max_reference_images
4
output_formats
  • png
  • jpeg
  • webp
defaults
size
1024x1024
quality
medium

Pricing Breakdown

TypeRate
Image$0.0454 / image
Input$5.40 / 1M tokens
Output$43.20 / 1M tokens
Image input$10.80 / 1M tokens
Low quality$0.0119 / image
Medium quality$0.0454 / image
High quality$0.1804 / image

image is medium-quality generation price per image

Supported Parameters

promptsizequalitybackgroundoutput_formatnuser

Specifications

Release date2025-04-23openai.com β†—verified
ArchitectureNatively multimodal language model for text and image inputs to image outputs; deeper architecture details not publicly discloseddevelopers.openai.com β†—verified
Native output resolutions1024Γ—1024, 1024Γ—1536, 1536Γ—1024developers.openai.com β†—verified
Output formatsPNG, JPEG, WebPdevelopers.openai.com β†—verified
Background supportTransparent, opaque, and auto; transparent output must use PNG or WebPdevelopers.openai.com β†—verified
Max images per call1 to 10 via the n parameterdevelopers.openai.com β†—verified
Quality tiersLow, medium, high, and auto for GPT Image modelsdevelopers.openai.com β†—verified
LicenseProprietary (OpenAI API terms)verified

Benchmarks

BenchmarkDistributionScoreSource
GIE-Bench
Highest functional-correctness among tested modelsarxiv.org β†—
STRICT text-rendering
One of only two proprietary models with low error rates on dense text
β€”Not publicly disclosedβ€”

API Usage Examples

Use the global api.therouter.ai endpoint shown below for new integrations; the legacy China accelerated endpoint is retired.

Recommended: use the async API

Image generation typically takes 30–180s, beyond the edge sync timeout. The examples below use the ?async=true submit + poll pattern. Read the full async image generation & edit guide β†’

cURL
# 1) Submit job (returns 202 immediately with a polling URL).
# Image generation takes 30-180s β€” always use the async path in production.
JOB=$(curl -s -X POST "https://api.therouter.ai/v1/images/generations?async=true"   -H "Content-Type: application/json"   -H "Authorization: Bearer $THE_ROUTER_API_KEY"   -d '{
    "model": "openai/gpt-image-1",
    "prompt": "A cinematic product render with soft studio lighting"
  }' | python3 -c "import sys,json;print(json.load(sys.stdin)['id'])")
echo "submitted: $JOB"

# 2) Poll until terminal (succeeded / failed / cancelled / expired).
while :; do
  R=$(curl -s "https://api.therouter.ai/v1/jobs/$JOB"     -H "Authorization: Bearer $THE_ROUTER_API_KEY")
  S=$(echo "$R" | python3 -c "import sys,json;print(json.load(sys.stdin)['status'])")
  echo "status: $S"
  case "$S" in
    succeeded) echo "$R" | python3 -c "import sys,json;print(json.load(sys.stdin)['unsigned_urls'][0])"; break ;;
    failed|cancelled|expired) echo "$R"; exit 1 ;;
  esac
  sleep 5
done

Image generation

Generate images from text prompts using the standard OpenAI Images API. Supports quality tiers, output format selection, transparent backgrounds, and batch generation.

cURL
curl https://api.therouter.ai/v1/images/generations \
  -H "Authorization: Bearer $THEROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/gpt-image-1",
    "prompt": "A serene mountain landscape at sunset, photorealistic style",
    "n": 1,
    "size": "1024x1024",
    "quality": "medium",
    "output_format": "png"
  }'

Image Editing Examples

Upload an image and describe the edit you want with a text prompt; the model returns the edited image as base64.

cURL
# Same async submit + poll pattern as /v1/images/generations.
JOB=$(curl -s -X POST "https://api.therouter.ai/v1/images/edits?async=true"   -H "Authorization: Bearer $THE_ROUTER_API_KEY"   -F "model=openai/gpt-image-1"   -F "prompt=Turn this scene into a watercolor painting"   -F "size=1024x1024"   -F "image=@input.png" | python3 -c "import sys,json;print(json.load(sys.stdin)['id'])")
echo "submitted: $JOB"

while :; do
  R=$(curl -s "https://api.therouter.ai/v1/jobs/$JOB"     -H "Authorization: Bearer $THE_ROUTER_API_KEY")
  S=$(echo "$R" | python3 -c "import sys,json;print(json.load(sys.stdin)['status'])")
  echo "status: $S"
  case "$S" in
    succeeded) echo "$R" | python3 -c "import sys,json;print(json.load(sys.stdin)['unsigned_urls'][0])"; break ;;
    failed|cancelled|expired) echo "$R"; exit 1 ;;
  esac
  sleep 5
done

More from openai

Similar models

Cross-provider sibling models

News & changes

2026-04-21

OpenAI launches ChatGPT Images 2.0 with gpt-image-2

OpenAI unveiled gpt-image-2, the next generation of its image model, with 2K resolution output, multilingual text rendering improvements, and agentic reasoning capabilities. The new model builds on gpt-image-1's foundation while significantly upgrading quality and feature set.

re-authored by TheRouteropenai.com β†—
2025-12-16

OpenAI releases gpt-image-1.5 with 4Γ— speed boost

OpenAI launched gpt-image-1.5, succeeding gpt-image-1 with approximately 4Γ— faster generation, improved instruction following, better text rendering for dense/small text, and more precise image editing with logo and face preservation.

re-authored by TheRoutertechcrunch.com β†—
2025-04-23

OpenAI brings gpt-image-1 to the API

OpenAI released gpt-image-1 to the API, bringing the native multimodal image generation that powered ChatGPT's viral image feature (130M+ users, 700M+ images in first week) to developers and enterprises through a standard Images API endpoint.

re-authored by TheRouteropenai.com β†—

Recent coverage

Frequently asked

How is gpt-image-1 different from DALLΒ·E 3?

gpt-image-1 is a natively multimodal autoregressive transformer, not a diffusion model. This means it has world-knowledge grounding from the GPT-4o family, can follow highly detailed instructions, renders text inside images more accurately, and supports image inputs for editing. DALLΒ·E 3 uses a diffusion-based approach and is typically faster for simple single-shot drafts at lower resolutions.

Can I use gpt-image-1 through TheRouter with the OpenAI SDK?

Yes. gpt-image-1 is fully compatible with the OpenAI Images API. Point your OpenAI SDK client to https://api.therouter.ai/v1, set the model to "openai/gpt-image-1", and use the images.generate() or /v1/images/generations endpoint. Gpt-image-1 uses the Images API, not the Chat Completions API.

Is there a cheaper alternative to gpt-image-1?

Yes. OpenAI offers gpt-image-1-mini, which costs approximately 80% less than gpt-image-1 while maintaining solid quality for most use cases. Additionally, gpt-image-1 has three quality tiers (low, medium, high) β€” choosing 'low' quality significantly reduces per-image cost for drafts and quick iterations.

re-authored by TheRouteren.wikipedia.org β†—
What safety features does gpt-image-1 have?

gpt-image-1 includes C2PA provenance metadata in all generated images, tunable content moderation via the moderation parameter (auto or low), and a safety stack that restricts harmful image generation. OpenAI does not train on customer API data by default.

Fact ledger β€” every claim on this page traces here
sourceURLretrieved
Release dateopenai.com β†—2026-05-29verified
Architecturedevelopers.openai.com β†—2026-08-11verified
Native output resolutionsdevelopers.openai.com β†—2026-08-11verified
Output formatsdevelopers.openai.com β†—2026-08-11verified
Background supportdevelopers.openai.com β†—2026-08-11verified
Max images per calldevelopers.openai.com β†—2026-08-11verified
Quality tiersdevelopers.openai.com β†—2026-08-11verified
Licenseβ€”β€”verified
GIE-Bencharxiv.org β†—2026-05-29verified
STRICT text-renderingarxiv.org β†—2026-05-29unknown
OpenAI launches ChatGPT Images 2.0 with gpt-image-2openai.com β†—2026-05-29verified
OpenAI releases gpt-image-1.5 with 4Γ— speed boosttechcrunch.com β†—2026-05-29verified
OpenAI brings gpt-image-1 to the APIopenai.com β†—2026-05-29verified
How is gpt-image-1 different from DALLΒ·E 3?openai.com β†—2026-05-29to verify
Is there a cheaper alternative to gpt-image-1?en.wikipedia.org β†—2026-05-29to verify
What safety features does gpt-image-1 have?openai.com β†—2026-05-29to verify
Help & contact