> ## 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.

# Is GPT Image or Nano Banana down? Live image model status and success rates

> Find out whether GPT Image or Nano Banana is down: the current status of each web, VIP, and Nano Banana model, success rates for the last hour and for each of the past 24 hours, and how to retry or switch models when a rate drops. Updated every 5 minutes.

The board below shows the combined success rate of all LaoZhang API requests to these image models, refreshed every 5 minutes. When a model turns yellow or red, retry or switch models as described in [What to do when the rate drops](#what-to-do-when-the-rate-drops). For the outcome of your own requests, check your [usage log](https://docs.laozhang.ai/en/faq/call-logs.md).

* **Models**: GPT Image web and VIP models, plus Nano Banana Pro, Nano Banana 2, and Nano Banana 2.1
* **Scope**: all keys and token groups combined, not per account
* **Times**: shown in your browser's time zone

## How to read the board

* **Status**: based on requests sent roughly 5 to 35 minutes ago. 95% or higher is Normal, 80% to 94% is Degraded, and below 80% is Down. Some requests from the last 5 minutes are still waiting for a result, so they are counted in the next update.
* **Last hour**: the success rate of requests sent in the hour before that 5-minute cutoff.
* **24-hour bar**: one block per hour, colored with the same thresholds. Gray means too few requests that hour to calculate a rate; a dashed outline marks hours before tracking started.
* **Below 95%**: hours under 95% are listed under each model. Select any block to see that hour's success rate and its most common errors.

## What to do when the rate drops

| Model | What to do |
| - | - |
| GPT Image VIP | Wait 30–60 seconds and retry once with the same model. If it fails again and `size` is 2K or smaller, resend with `model` set to `gpt-image-2.6-web`. For 4K, keep retrying VIP or use official forwarding |
| GPT Image web route | Wait 30–60 seconds and retry once. If it fails again, switch to `gpt-image-2.5-sunburst-vip` or `gpt-image-2.6-sunburst-vip` |
| Nano Banana | Wait 30–60 seconds, retry once, and lower your concurrency. If requests keep failing, find the request ID in your usage log and open a ticket |

* **Price**: the VIP and web route models all cost \$0.03 per image, so switching does not change what you pay.
* **Token groups**: a `default` group key can call both VIP and web route models. A `GPT Image VIP` group key can only call VIP models.
* **Automatic fallback**: for complete code that resends to the web route when VIP fails, see [Fall back to the web route when VIP fails](https://docs.laozhang.ai/en/api-capabilities/gpt-image-2-6.md#fall-back-to-the-web-route-when-vip-fails).

## How rates are calculated

* **Success**: the request returned HTTP 200.
* **Failure**: the request returned 429 or a 5xx error such as 500, 502, 503, or 504.
* **Not counted**: other 4xx errors such as invalid parameters, authentication failures, and moderation rejections, plus requests the client closed before the result arrived. These depend on the request itself and say nothing about whether the model is available.
* **One row per product**: a product's release ID and preview ID share a row. For example, the Nano Banana Pro row covers `gemini-3-pro-image` and `gemini-3-pro-image-preview`.

## Get help

* **Check your own requests**: filter your [usage log](https://docs.laozhang.ai/en/faq/call-logs.md) by model and time to see each request's result and charge.
* **Requests keep failing**: note the request ID from the usage log and open a ticket in the [console](https://api.laozhang.ai). We check each one.
* **Fall back in code**: to resend to the web route when VIP fails, see [Fall back to the web route when VIP fails](https://docs.laozhang.ai/en/api-capabilities/gpt-image-2-6.md#fall-back-to-the-web-route-when-vip-fails). For timeouts, see [Request timeouts](https://docs.laozhang.ai/en/faq/request-timeout.md); for 429 errors and concurrency, see [Concurrency and 429](https://docs.laozhang.ai/en/faq/api-concurrency.md).
* **Billing questions**: see [Image API billing and failure rules](https://docs.laozhang.ai/en/api-capabilities/image-billing-and-failures.md).
* **Incident notices**: incidents with a wide impact are posted in [Announcements](https://docs.laozhang.ai/en/changelog.md). You can also follow the [Telegram channel](https://t.me/laozhang_ai).

## FAQ

### Is Nano Banana 2 down right now?

Check the Nano Banana 2 row on this page. A Normal status and green blocks mean requests through LaoZhang API are returning images. Yellow or red means the success rate is dropping, so follow [What to do when the rate drops](#what-to-do-when-the-rate-drops). The row covers both `gemini-3.1-flash-image` and `gemini-3.1-flash-image-preview`; Nano Banana Pro and 2.1 have rows of their own.

### Is GPT Image 2 stable right now?

Check the status and last-hour rate in the `gpt-image-2-vip` row. When VIP is unstable, send 1K and 2K images to `gpt-image-2.6-web` instead, at the same \$0.03 per image. The `gpt-image-2.6-web` row shows how the web route is doing.

### How is this different from the OpenAI or Google status page?

An official status page tells you whether the provider's own service has an incident. This page measures requests actually sent through LaoZhang API, including our gateway and routes, so the two don't always move together. To decide whether a model works on LaoZhang API, go by this page and your [usage log](https://docs.laozhang.ai/en/faq/call-logs.md).

### Why is my own success rate different from the board?

The board combines every user's requests. Your own results also depend on your parameters, such as size, quality, reference image size, and concurrency. Filter your [usage log](https://docs.laozhang.ai/en/faq/call-logs.md) by model and time to see each of your requests.

### My model isn't on the board. Where do I check it?

The board lists only models with enough steady traffic to calculate a reliable rate. For other models, check your own requests in the usage log. Outages with a wide impact are posted in [Announcements](https://docs.laozhang.ai/en/changelog.md).

### Am I charged for failed requests while the rate is low?

No. Every request the board counts as failed returned 429 or a 5xx error, and requests that get a non-200 response from upstream are not charged. One exception applies to Nano Banana: when Google's moderation blocks a prompt, it returns HTTP 200, so the request is charged once and counts as a success here. See [Image API billing and failure rules](https://docs.laozhang.ai/en/api-capabilities/image-billing-and-failures.md) for the full rules.

### The board says Normal but my requests fail. What now?

Check the error code first. 400, 401, 403, and moderation rejections come from the request itself, so fix it according to the error message and send it again. For 429 and 5xx errors, retry after 30–60 seconds. If requests keep failing, find the request ID in your usage log and open a ticket.

## Related documentation

* [GPT Image 2.6 guide](https://docs.laozhang.ai/en/api-capabilities/gpt-image-2-6.md): web route, VIP, and official-forwarding parameters, plus fallback code
* [Image API billing and failure rules](https://docs.laozhang.ai/en/api-capabilities/image-billing-and-failures.md): which failures are charged, timeouts, and retries
* [Nano Banana 2.1 guide](https://docs.laozhang.ai/en/api-capabilities/nano-banana-2-1-image.md)
* [Models and pricing](https://docs.laozhang.ai/en/models.md), [usage log](https://docs.laozhang.ai/en/faq/call-logs.md)


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.