# GPT Image 2.5 Sunburst API pricing comparison (October 2026)

By SeedRouter · Published 2026-10-21 · Updated 2026-10-21

GPT Image 2.5 Sunburst API pricing has two useful units: a price per delivered image and a price per million tokens. OpenAI and fal bill Sunburst at **$30 per million image output tokens**, so one 1K image at `high` quality comes to about $0.053; WaveSpeedAI lists $0.09 and Replicate $0.128 for the same quality. At the prices checked on October 21, 2026, **SeedRouter's Standard price per image was at least 57% lower** than each of them at `high`, `xhigh` and `max` quality. \[1]\[2]\[3]\[4]\[5]

SeedRouter's current Sunburst prices appear below, using the same live rate source and refresh interval as the [model pricing calculator](https://seedrouter.ai/models/gpt-image-2-5#pricing). You can [try GPT Image 2.5 in the playground](https://seedrouter.ai/models/gpt-image-2-5) at the same rates.

## What does Sunburst cost on SeedRouter today?

GPT Image 2.5 Sunburst prices are listed separately for each model ID. Per-image prices follow the selected quality; token-billed prices follow reported usage.

| Channel | Model IDs | Billed by | Current rate (USD) |
| --- | --- | --- | --- |
| Standard | `gpt-image-2.5-sunburst` | Per delivered image at the selected model and quality rate | Per delivered image: xhigh $0.0456; max $0.0456; other quality values $0.036 |
| Official | `gpt-image-2.5-sunburst-official` | Tokens each render reports (text input + image output) | $3.75 per 1M input tokens; $22.5 per 1M output tokens |

Token rates are the published first-tier rates; usage in a different tier may be charged differently.

### Cost of one 1024x1024 image by quality

Token counts are measured samples on the Official channel, with 18 input tokens for the prompt. Actual usage varies with the prompt, dimensions, references, and image count, so token-billed figures are estimates, not fixed prices. Per-image rates below follow each model's current quality pricing.

| Quality | Sample output tokens | Standard USD | Official USD |
| --- | --- | --- | --- |
| low | 196 | $0.036 | $0.0044775 |
| medium | 439 | $0.036 | $0.009945 |
| high | 1756 | $0.036 | $0.0395775 |
| xhigh | 3122 | $0.0456 | $0.0703125 |
| max | 7024 | $0.0456 | $0.1581075 |
| auto | 229 | $0.036 | $0.00522 |

Formula (Official): (input tokens × 3.75 + output tokens × 22.5) / 1,000,000 USD.
A request for n images is billed for the images delivered. Final charges are available in account usage history. Failed requests are not charged.

Standard bills per delivered image at the selected quality rate. Official bills the token usage a render reports. The quality examples use the model page's measured samples; they are useful for budgeting, but they are not guaranteed token counts for every prompt.

The table covers Sunburst specifically. For the other member of the family, see the [Flare and Sunburst comparison](https://seedrouter.ai/blog/gpt-image-2-5-flare-vs-sunburst). A cheaper Flare quote should not be presented as a Sunburst price.

## How much less does SeedRouter cost than OpenAI and other APIs?

USD per 1K GPT Image 2.5 Sunburst image, by quality setting, using SeedRouter’s Standard model ID.

| Platform | high | xhigh | max |
| --- | --- | --- | --- |
| SeedRouter (live) | $0.036 | $0.0456 | $0.0456 |
| [Replicate](https://replicate.com/openai/gpt-image-2.5-sunburst) | $0.128 (SeedRouter 72% less) | $0.25 (SeedRouter 82% less) | $0.50 (SeedRouter 91% less) |
| [WaveSpeedAI](https://wavespeed.ai/models/openai/gpt-image-2.5-sunburst/text-to-image) | $0.09 (SeedRouter 60% less) | $0.16 (SeedRouter 72% less) | $0.36 (SeedRouter 87% less) |
| [Atlas Cloud](https://www.atlascloud.ai/pricing) | $0.0563 (SeedRouter 36% less) | $0.10 (SeedRouter 54% less) | $0.225 (SeedRouter 80% less) |
| [OpenAI API](https://developers.openai.com/api/docs/models/gpt-image-2.5-sunburst) | $0.0528 (SeedRouter 32% less) | $0.0938 (SeedRouter 51% less) | $0.211 (SeedRouter 78% less) |
| [fal](https://fal.ai/models/openai/gpt-image-2.5/sunburst/text-to-image) | $0.0528 (SeedRouter 32% less) | $0.0938 (SeedRouter 51% less) | $0.211 (SeedRouter 78% less) |

Other platforms' prices were checked on October 21, 2026 and may have changed since. SeedRouter's prices are read live from the rates that bill your account. OpenAI, fal and Atlas Cloud bill output tokens ($30, $30 and $32 per million); their figures use the output tokens SeedRouter measured for one 1K image at each quality (1,756, 3,122 and 7,024), so they are estimates. Replicate and WaveSpeedAI publish a price per image.

The SeedRouter row is read from the same rates that bill your account, so it changes when SeedRouter's prices change. The other rows are a dated snapshot of each platform's own page, without promotional credits or payment fees.

fal also charges $8 per million image input tokens and $5 per million text input tokens, and rounds each total up to $0.0001. Atlas Cloud bills output tokens at $32 per million. fal's documented default quality is `high`. If a comparison request omits quality on one service but sets `medium` on another, the result may be a comparison of different work rather than different prices. Set the parameters explicitly. \[2]\[3]

## Why is there no universal price for a 1K image?

Dimensions alone do not fix a token-based bill. Quality, input images and the actual output usage also matter. A reference-heavy editing request has an input cost that a text-only example does not capture. A per-image price, such as SeedRouter Standard, avoids that variation: the bill depends on the selected quality, not on how many tokens a render reports.

For a practical comparison, write down the full request: one image, explicit dimensions such as `1024x1024`, an explicit quality, and either no reference image or the same reference input. Compare actual reported usage after a successful request with the applicable rate. For fixed billing, compare the selected model and quality's price per delivered image.

Size labels also need care. Two services can use “2K” for different pixel dimensions. Check the actual width and height instead of assuming the labels describe an identical canvas.

## How should you budget a batch of images?

Start with the number of successful outputs you expect to generate, including candidates you will discard. A team needing 100 finished product images and expecting three candidates per final should budget for 300 delivered images.

For fixed billing, use the current unit rate at each quality:

```text
budget = draft images × current draft price
       + final images × current final price
```

For token billing, estimate each separately billed category:

```text
budget = sum(category tokens × category USD per million) / 1,000,000
```

Use the categories the service actually bills; fal, for example, lists separate text-input, image-input and image-output rates. Do not add a separate category to an estimate if it is already included in another reported charge. Keep cached input separate from ordinary input when applicable. \[2]

A useful production split is 200 draft candidates and 100 final outputs. Apply the draft quality's rate to the first group and the final quality's rate to the second. Multiplying the lowest advertised rate by all 300 images can understate the budget when the finals use a more expensive quality.

For the broader billing walkthrough, see [GPT Image 2.5 cost per image](https://seedrouter.ai/blog/gpt-image-2-5-api-pricing). That guide covers the family; this comparison focuses on Sunburst and other platforms' list prices.

## What matters when choosing a Sunburst API?

A clear price is only useful if the service supports the request you need. Confirm the model name, quality values, dimensions and editing inputs. Read the request contract before moving a working integration between services: similar model names do not guarantee identical request bodies or response formats.

On SeedRouter, start with the [GPT Image 2.5 API guide](https://seedrouter.ai/docs/gpt-image-2-5) and use the model ID shown in the live table. Failed requests are not charged, the balance is prepaid with no subscription, and credits never expire. A successful image that does not fit your creative brief still counts as a successful generation, so candidate selection belongs in the budget.

Use the final charge in [usage history](https://seedrouter.ai/usage) to refine the next estimate. A small representative workload is more informative than a single minimum-price example, especially for edits with several reference images.

## Frequently asked questions

### Is GPT Image 2.5 Sunburst billed per image or per token?

Both billing methods exist. On SeedRouter, Standard bills delivered images at the selected quality rate and Official bills reported token usage. OpenAI, fal and Atlas Cloud bill tokens; WaveSpeedAI and Replicate price each image. Compare the unit as well as the amount.

### Does high quality always have the same price as low quality?

No. Check the exact platform and model ID. Token usage rises with quality, and a per-image service can also use quality-specific rates. The current table shows the applicable SeedRouter quality rates.

### Is Sunburst cheaper on SeedRouter than on OpenAI's API?

At `high`, `xhigh` and `max` quality, yes: at the October 21, 2026 prices SeedRouter Standard was at least 57% lower per 1K image than OpenAI, fal, Atlas Cloud, WaveSpeedAI and Replicate. The comparison table recalculates the difference from SeedRouter's live rates.

### Will the prices in this article change automatically?

The SeedRouter tables use the same live prices and refresh interval as the model page. Other platforms' quotes remain tied to their stated check date. If current rates cannot load, the table reports that they are unavailable; an unavailable rate is not a free rate.

## Sources

1. [OpenAI: GPT Image 2.5 Sunburst model and token pricing](https://developers.openai.com/api/docs/models/gpt-image-2.5-sunburst).
2. [fal: Sunburst pricing, rounding and default quality](https://fal.ai/models/openai/gpt-image-2.5/sunburst/text-to-image).
3. [Atlas Cloud pricing](https://www.atlascloud.ai/pricing).
4. [WaveSpeedAI: GPT Image 2.5 Sunburst](https://wavespeed.ai/models/openai/gpt-image-2.5-sunburst/text-to-image).
5. [Replicate: GPT Image 2.5 Sunburst](https://replicate.com/openai/gpt-image-2.5-sunburst).