GPT Image 2.5 VIP outage on September 16

This incident is resolved. Both gpt-image-2.5-flare-vip and gpt-image-2.5-sunburst-vip are available again as of September 22, 2026. Read the September 22 recovery announcement for current status and steps to resume requests.

• Published: September 16, 2026
• Last checked: September 22, 2026 (recovery status)
• Current status: Restored; this page preserves the September 16 incident and the alternatives recommended at the time

The record below describes the September 16 outage. Its advice to pause requests or switch routes no longer applies.

​

Models and alternatives during the outage

This incident affected only the two GPT Image 2.5 -vip models listed above. Users of official-forward models, gpt-image-2.5-web, and gpt-image-2-vip did not need to switch routes because of this notice.

​

Guidance issued during the outage (historical)

1. Pause new requests to both 2.5 -vip models and move jobs that need to continue to an available route.
2. For explicit Flare / Sunburst selection or 2K / 4K output: choose an unsuffixed official-forward model. Check official-forward permissions and usage-based billing in token management, then follow the GPT Image 2.5 integration guide.
3. For GPT Image 2.5 with per-call billing: use gpt-image-2.5-web. It accepts size, but cannot select Flare / Sunburst. Do not assume that every parameter from the two -vip models carries over; verify the parameters and output you need with your current account.
4. If GPT Image 2 is sufficient: use gpt-image-2-vip and follow the GPT Image 2 guide.

Before resuming batch jobs, run a small-volume request, open the resulting image, check its dimensions, and review the model, usage, and charge in call logs. Official forwarding uses usage-based billing; check the console model and pricing page before switching.

​

Incident and recovery questions

​

Did the incident affect all GPT Image 2.5 access?

No. Only gpt-image-2.5-flare-vip and gpt-image-2.5-sunburst-vip were affected. The unsuffixed official-forward models and gpt-image-2.5-web remained available.

​

Can the web route select Flare or Sunburst?

No. gpt-image-2.5-web supports size, but cannot manually select Flare / Sunburst. Use the corresponding official-forward model if you need a specific variant.

​

Will the replacement still use per-call billing?

gpt-image-2.5-web and gpt-image-2-vip use per-call billing. Official-forward gpt-image-2.5-flare / gpt-image-2.5-sunburst use usage-based billing. Check both the token configuration and the target model’s price.

​

Have both 2.5 -vip models returned?

Both models are available again as of September 22, 2026. See the September 22 recovery announcement.

​

Sources and related links

• LaoZhang API route status: this notice reflects the September 16, 2026 operations update on availability, billing types, and parameter scope. It is not a per-account test of every parameter combination. Verify current account access, actual output, and charges through the console and call logs.
• Upstream API documentation: the OpenAI image generation guide provides background for official forwarding; this announcement is the source for LaoZhang API -vip / -web route status.
• Related documentation: GPT Image 2.5 integration guide, GPT Image 2.5 launch announcement, and service and billing terms.