> ## Documentation Index
> Fetch the complete documentation index at: https://docs.bytespike.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# GPT-Image 2

> OpenAI's photorealistic image generator — strong on prompts that mix description with explicit composition rules.

**Vendor:** OpenAI
**Model ID:** `gpt-image-2`
**Capability:** 1024² – 4096² · photorealism · prompt-following · text-in-image
**Pricing:** per image, photoreal tier ([live rate](https://bytespike.ai/pricing#image))

GPT-Image 2 is OpenAI's photoreal generator. The thing that
distinguishes it from Nano Banana is *prompt following* — when your
brief reads "the kettle on the LEFT, mug on the RIGHT, steam rising
from both", GPT-Image 2 respects spatial relationships better than
most photoreal alternatives. For pure material fidelity, Nano Banana
Pro and 2 still have an edge; for instruction adherence, this is
usually the right call.

## Request

```bash theme={null}
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 kettle on the left and a mug on the right, both steaming, on a marble counter",
    "size": "1024x1024",
    "quality": "high"
  }'
```

### Body parameters

| Field                | Type    | Required | Default     | Notes                                                                       |
| -------------------- | ------- | -------- | ----------- | --------------------------------------------------------------------------- |
| `model`              | string  | yes      | —           | `gpt-image-2`                                                               |
| `prompt`             | string  | yes      | —           | English-tuned.                                                              |
| `size`               | string  | no       | `1024x1024` | Supported: `1024x1024`, `1024x1536`, `1536x1024`, `2048x2048`, `4096x4096`. |
| `n`                  | integer | no       | 1           | 1–4 images.                                                                 |
| `quality`            | string  | no       | `"medium"`  | `"low"` / `"medium"` / `"high"` / `"auto"`.                                 |
| `style`              | string  | no       | `"natural"` | `"natural"` / `"vivid"`.                                                    |
| `output_format`      | string  | no       | `"png"`     | `"png"` / `"webp"` / `"jpeg"`.                                              |
| `output_compression` | integer | no       | —           | 0–100, only for `webp` / `jpeg`.                                            |

## Response

```json theme={null}
{
  "created": 1715174400,
  "data": [
    {
      "url": "https://cdn.bytespike.ai/img/...",
      "revised_prompt": "..."
    }
  ]
}
```

URLs pre-signed, 24h expiry.

## Code examples

<CodeGroup>
  ```bash cURL theme={null}
  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": "Kettle and mug...", "quality": "high"}'
  ```

  ```python Python theme={null}
  from openai import OpenAI

  client = OpenAI(base_url="https://llm.bytespike.ai/v1", api_key="$BYTESPIKE_API_KEY")
  resp = client.images.generate(
      model="gpt-image-2",
      prompt="A kettle on the left and a mug on the right, both steaming, on a marble counter",
      size="1024x1024",
      quality="high",
  )
  print(resp.data[0].url)
  ```

  ```javascript Node theme={null}
  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 kettle on the left and a mug on the right, both steaming, on a marble counter",
    size: "1024x1024",
    quality: "high",
  })
  console.log(resp.data[0].url)
  ```
</CodeGroup>

## Errors

| Code                  | Trigger                           | Billed?         |
| --------------------- | --------------------------------- | --------------- |
| 400 / 401 / 402 / 403 | Standard                          | No              |
| 451                   | Prompt blocked by upstream safety | No              |
| 5xx                   | Upstream issue                    | No (auto-retry) |

## When to use

* Prompts that specify spatial relationships, counts, or composition.
* Briefs that mix text descriptions with explicit constraints.
* For more material / lighting fidelity, see [Nano Banana 2](/api-reference/image/nano-banana-v2) or [Nano Banana Pro](/api-reference/image/nano-banana-pro).
* For the compliance-tier variant, see [GPT-Image 2 Official](/api-reference/image/gpt-image-2-official).

## Limits

| Limit                         | Value                                      |
| ----------------------------- | ------------------------------------------ |
| Max output resolution         | 4096×4096                                  |
| Max images per call (`n`)     | 4                                          |
| Supports init image (img2img) | No                                         |
| Supports `quality` modifier   | Yes                                        |
| Supports `style` modifier     | Yes                                        |
| Sync?                         | Yes (≤30s typical, longer at high quality) |
| Avg latency for 1024² high    | 12-18s                                     |