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

# GPT Image 2.5 API: Official Forwarding and Per-Call Billing

> Integrate GPT Image 2.5 Flare and Sunburst through token-billed official forwarding or $0.03/call routes, with model IDs, groups, generation and editing examples, and web version 2.5 guidance.

`gpt-image-2.5-flare` and `gpt-image-2.5-sunburst` are available through official API forwarding, using the same Images API integration as official-forward `gpt-image-2`. Choose **Flare** for speed or **Sunburst** for precise image editing, or expose both options in your app. Both use token billing at the same token rates as official-forward `gpt-image-2`.

Per-call billing is available with `gpt-image-2.5-flare-vip` and `gpt-image-2.5-sunburst-vip`, each at **\$0.03/call**, using the same integration as `gpt-image-2-vip`.

**Updated September 9, 2026.**

## Choose your route

| Route               | Model ID                     | Token group                                              | Billing                                            |
| ------------------- | ---------------------------- | -------------------------------------------------------- | -------------------------------------------------- |
| Official forwarding | `gpt-image-2.5-flare`        | `Sora2Official` or the enterprise official-forward group | Same token rates as official-forward GPT Image 2   |
| Official forwarding | `gpt-image-2.5-sunburst`     | `Sora2Official` or the enterprise official-forward group | Same token rates as official-forward GPT Image 2   |
| Per-call billing    | `gpt-image-2.5-flare-vip`    | Default                                                  | \$0.03/call, same integration as `gpt-image-2-vip` |
| Per-call billing    | `gpt-image-2.5-sunburst-vip` | Default                                                  | \$0.03/call, same integration as `gpt-image-2-vip` |
| Web version 2.5     | `gpt-image-2-web`            | Default                                                  | Currently \$0.03/call; check the console           |

Choose the route and billing type before copying a request. Existing GPT Image 2 integrations can refer to the [GPT Image 2 guide](/en/api-capabilities/gpt-image-2).

## GPT Image 2.5 official-forward models and groups

| Request model ID         | Current snapshot                    | When to choose it                                        |
| ------------------------ | ----------------------------------- | -------------------------------------------------------- |
| `gpt-image-2.5-flare`    | `gpt-image-2.5-flare-2026-09-08`    | Faster everyday, high-quality generation                 |
| `gpt-image-2.5-sunburst` | `gpt-image-2.5-sunburst-2026-09-08` | Generation and editing where edit precision matters most |

Use an undated alias for a typical integration, or the full dated ID to pin a snapshot. The current alias mapping may change in the future.

1. Select usage-based billing in [token management](https://api2.laozhang.ai/token) and confirm that the token allows your target model.
2. Both 2.5 models require an **official-forward group**. Choose `Sora2Official` for regular official forwarding or the `GPTImage2 Sora2 Enterprise` official-forward group when stability is the priority. **Default continues to use `gpt-image-2-web`, which now runs the latest GPT web version, 2.5.**
3. Keep your existing official-forward Images API authentication, request structure, and result handling. Change `model` to an alias or snapshot above; do not put a group name in `model`.

<Note>
  The console groups for official forwarding are `Sora2Official` and `GPTImage2 Sora2 Enterprise`; their configuration keys are `sora_official` and `GPT_Image_2_Enterprise`. Select the group in token settings, not in the request’s `model` field. For per-call access through Default, use a model ID ending in `-vip`.
</Note>

## Minimal GPT Image 2.5 official-forward integration

These examples use the existing official-forward endpoints: `/v1/images/generations` for generation and `/v1/images/edits` for editing. The SDK methods are `images.generate` and `images.edit`.

```bash theme={null}
export LAOZHANG_API_KEY="sk-your-usage-billed-token"
export BASE_URL="https://api2.laozhang.ai/v1"
```

### Generate with Flare

```bash theme={null}
curl --fail-with-body "$BASE_URL/images/generations" \
  -H "Authorization: Bearer $LAOZHANG_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-image-2.5-flare",
    "prompt": "A product photo of a white ceramic mug on a gray desk in soft natural light",
    "size": "1024x1024",
    "quality": "medium"
  }' -o generation.json
```

### Edit with Sunburst

Prepare a local `source.png`, then run:

```bash theme={null}
curl --fail-with-body "$BASE_URL/images/edits" \
  -H "Authorization: Bearer $LAOZHANG_API_KEY" \
  -F "model=gpt-image-2.5-sunburst" \
  -F "image=@source.png" \
  -F "prompt=Keep the mug shape, composition, and lighting. Change only the mug body to dark blue." \
  -F "size=1024x1024" \
  -F "quality=medium" \
  -o edit.json
```

Both models accept text and image inputs. These examples reflect their respective strengths; Flare can also edit, and Sunburst can also generate.

<Info>
  Refer to the [OpenAI image generation guide](https://developers.openai.com/api/docs/guides/image-generation) for official-forward API parameters. The examples on this page use Images API: `/v1/images/generations` for generation and `/v1/images/edits` for editing. Per-call quality options are listed under “Size and quality” below; keep the parameter ranges for the two routes separate.
</Info>

### Save and inspect the image

A successful Images API response should contain an image result. For `data[0].b64_json`, save the generation result with this script; read `edit.json` for an edit request:

```python theme={null}
import base64
import json
from pathlib import Path

response = json.loads(Path("generation.json").read_text())
value = response["data"][0]["b64_json"]
if value.startswith("data:"):
    value = value.split(",", 1)[1]
Path("output.png").write_bytes(base64.b64decode(value, validate=True))
```

Open `output.png` and check that it displays correctly at the requested dimensions. For edits, inspect both the requested change and the details that should remain intact. If the route returns `data[0].url`, download and open that image. Then check the model, group, usage, and actual charge in [call logs](https://api2.laozhang.ai/log).

For authentication or model-access errors, check the key, group, and model ID first. Correct invalid parameters using the returned error. If no image is present, retain the sanitized error and log time, check the [Images API reference](/en/api-reference/images), and contact support without sharing a complete key.

## GPT Image 2.5 official-forward billing

Both 2.5 models have the **same token rates** as official-forward `gpt-image-2`. Billing covers actual text input, image input, and image output usage. The fixed per-call price of older routes does not apply.

| Official billing item | USD per 1 million tokens |
| --------------------- | ------------------------ |
| Text input            | \$5                      |
| Cached text input     | \$1.25                   |
| Image input           | \$8                      |
| Cached image input    | \$2                      |
| Image output          | \$30                     |

These are [OpenAI's official token rates](https://developers.openai.com/api/docs/guides/image-generation#cost-and-latency). Confirm LaoZhang API settlement against the [current model pricing and token group](https://api2.laozhang.ai/account/pricing); contact support for enterprise quotes. Equal token rates do not mean equal cost per image. The model, size, quality, and reference images affect usage, so estimate costs from representative requests.

## GPT Image 2.5 per-call integration

`gpt-image-2.5-flare-vip` and `gpt-image-2.5-sunburst-vip` are now available at **\$0.03/call** each, using the same integration and billing method as `gpt-image-2-vip`. Use a **Default-group per-call token** and confirm access to the target model.

| Model ID                     | Billing     | Integration               |
| ---------------------------- | ----------- | ------------------------- |
| `gpt-image-2.5-flare-vip`    | \$0.03/call | Same as `gpt-image-2-vip` |
| `gpt-image-2.5-sunburst-vip` | \$0.03/call | Same as `gpt-image-2-vip` |

### Size and quality

Both per-call models accept `size` and `quality` for image generation. Use **`low`, `medium`, or `high`** for `quality`. **Per-call routes currently reject `xhigh` and `max` with HTTP 400.** Omit either field when you do not need to choose its value.

| Parameter | Value                               | Usage                                                                                          |
| --------- | ----------------------------------- | ---------------------------------------------------------------------------------------------- |
| `size`    | `WIDTHxHEIGHT`, such as `2048x2048` | Set it explicitly when dimensions matter; choose a common size below                           |
| `quality` | `low` / `medium` / `high`           | Both models support these three tiers; select one for your task or expose the options to users |

Both per-call models support these common sizes. Choose a value for your required aspect ratio:

| Size category | `size`      | Aspect ratio |
| ------------- | ----------- | ------------ |
| 1K square     | `1024x1024` | 1:1          |
| Landscape     | `1536x1024` | 3:2          |
| Portrait      | `1024x1536` | 2:3          |
| 2K square     | `2048x2048` | 1:1          |
| 4K landscape  | `3840x2160` | 16:9         |
| 4K portrait   | `2160x3840` | 9:16         |

<Tip>
  Start with `size="2048x2048"` and `quality="medium"`. The quality setting controls generation quality while both models remain billed per call; check actual rates and charges in the [console](https://api2.laozhang.ai/account/pricing).
</Tip>

**Dimension handling:** Nonstandard sizes may be adjusted automatically; for example, `1023x1024` may return `1024x1024`. If exact pixels matter, use a standard size from the table and check the returned image dimensions.

### Generation example

For an existing `gpt-image-2-vip` integration, keep the base URL, authentication, endpoint, and result handling. Replace `model` with a full ID above. This example generates a 2K square image with `quality="medium"`:

```bash theme={null}
export LAOZHANG_API_KEY="sk-your-per-call-token"

curl --fail-with-body "https://api2.laozhang.ai/v1/images/generations" \
  -H "Authorization: Bearer $LAOZHANG_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-image-2.5-flare-vip",
    "prompt": "A product photo of a white ceramic mug on a gray desk in soft natural light",
    "size": "2048x2048",
    "quality": "medium"
  }' -o generation.json
```

### Editing example

For editing, use `/v1/images/edits` with the same multipart request format as `gpt-image-2-vip`. Prepare a local `source.png`, upload it, and describe the changes you want:

```bash theme={null}
curl --fail-with-body "https://api2.laozhang.ai/v1/images/edits" \
  -H "Authorization: Bearer $LAOZHANG_API_KEY" \
  -F "model=gpt-image-2.5-sunburst-vip" \
  -F "image=@source.png" \
  -F "prompt=Keep the mug shape, composition, and lighting. Change only the mug body to dark blue." \
  -o edit.json
```

Both per-call models share this integration method. Choose either model or offer users a selector. Save the image as shown above, and confirm actual prices and charges in the [console](https://api2.laozhang.ai/account/pricing) and call logs.

<Info>
  Keep the `-vip` suffix for per-call requests, such as `gpt-image-2.5-flare-vip`. Unsuffixed Flare / Sunburst models use official-forward token billing. For per-call quality settings, use `low`, `medium`, or `high`.
</Info>

## Default group: gpt-image-2-web now runs web version 2.5

The Default web route continues to use `gpt-image-2-web`. It runs the latest GPT web version, 2.5. Existing integrations can keep this model ID; the web update does not require switching to Flare or Sunburst.

`gpt-image-2-web` is a LaoZhang API web-route model name. It is a different integration route from Flare / Sunburst; web version 2.5 should not be equated with a specific official-forward snapshot. The current public catalog lists **\$0.03/call**, without a switch to official-forward token billing. Confirm actual charges and parameter limits for this route in the console.

## Related documentation

* [GPT Image 2: existing integrations](/en/api-capabilities/gpt-image-2)
* [Images API reference](/en/api-reference/images)
* [Image generation selection guide](/en/api-capabilities/image-generation-guide)
* [Call logs](/en/faq/call-logs)
* [Model and pricing catalog](/en/models)

## References

* [OpenAI Image generation](https://developers.openai.com/api/docs/guides/image-generation)
* [OpenAI GPT Image 2.5 Flare](https://developers.openai.com/api/docs/models/gpt-image-2.5-flare)
* [OpenAI GPT Image 2.5 Sunburst](https://developers.openai.com/api/docs/models/gpt-image-2.5-sunburst)
* [Current LaoZhang API models and pricing](https://api2.laozhang.ai/account/pricing)
* [GPT Image 2.5 launch notice](/en/announcements/gpt-image-2-5-2026-09)