gpt-image-2

gpt-image-2 is OpenAI’s flagship image model. Text-to-image and image-to-image (mask-driven edits) live on a single endpoint, billed per image generated. Strong on layout fidelity, in-image typography, and product/scene composition — the place to reach for when “looks like a real OpenAI render” matters more than the lowest per-image cost. Pricing:

0.08 / image — see the [rate card](/pricing). The gateway bills a flat

0.08 per generated image; failures don’t bill.

Protocols

Protocol	Path
OpenAI Images	`POST https://llm.bytespike.ai/v1/images/generations`

Quickstart

curl https://llm.bytespike.ai/v1/images/generations \
  -H "Authorization: Bearer $BYTESPIKE_API_KEY" \
  -H "content-type: application/json" \
  -d '{
    "model": "gpt-image-2",
    "prompt": "A flat-style red apple on a soft beige background, marketing hero",
    "n": 1,
    "size": "1024x1024"
  }'

from openai import OpenAI

client = OpenAI(
    base_url="https://llm.bytespike.ai/v1",
    api_key=os.environ["BYTESPIKE_API_KEY"],
)

resp = client.images.generate(
    model="gpt-image-2",
    prompt="A flat-style red apple on a soft beige background, marketing hero",
    n=1,
    size="1024x1024",
)

print(resp.data[0].url)

import OpenAI from "openai"

const client = new OpenAI({
  baseURL: "https://llm.bytespike.ai/v1",
  apiKey: process.env.BYTESPIKE_API_KEY,
})

const resp = await client.images.generate({
  model: "gpt-image-2",
  prompt: "A flat-style red apple on a soft beige background, marketing hero",
  n: 1,
  size: "1024x1024",
})

console.log(resp.data[0].url)

Capabilities

Capability	Supported
Text-to-image	✅
Image-to-image (with `source_image`)	✅
Mask-driven edits	✅
`n` ≥ 2 batch generation	✅
`size` 1024×1024 / 1024×1792 / 1792×1024	✅
`quality` standard / hd	✅
`response_format` url / b64_json	✅
Modality	image
Capability bucket	`image_generate`

When to use

Marketing creative — hero images, social cards, anywhere typography-in-image matters.
Product mockups — fidelity on materials, lighting, and small print holds up better than most domestic alternatives.
DOSIA generate_image tool — the main brain will resolve “draw me an X” to this model by default when permission is granted.

When not to use:

High-volume or budget-sensitive work — nano-banana-v2 is materially cheaper for the same shape.

nano-banana-v2 — cheaper Google image model with image-to-image
Multimodal endpoints — overview of image / video / audio / embedding surfaces

Get started

Client setup

Migrate from

Concepts

Models

Internal (DOSIA-coupled)

Protocols

Quickstart

Capabilities

When to use

Next

​Protocols

​Quickstart

​Capabilities

​When to use

​Next

Protocols

Quickstart

Capabilities

When to use

Next