Fix GPT Image 2.6 `GIF output is not enabled for this request` on Codex OAuth

TL;DR: GIF output is not enabled for this request on GPT Image 2.6 via Codex OAuth is an upstream GIF entitlement limitation, not a missing proxy parameter, so retrying the same request will not fix it. This guide is based on maintainer confirmation in router-for-me/CLIProxyAPI#7269, a collaborator investigation in kiss-kedaya/sub2api#195, and the OpenAI Images API reference cited in those reports.

Applies to: GPT Image 2.6 Flare and Sunburst GIF requests through Codex OAuth, including CLIProxyAPI v8.0.56 on Linux amd64 and sub2api-derived deployments reported on 2026-11-05. Last checked: 2026-11-06.

Related guides: GIF tutorial, GIF parameter errors and GIF pricing.

GIF output is not enabled for this request on generations

HTTP 400
GIF output is not enabled for this request.

GPT Image 2.6 shows GIF output is not enabled for this request on POST /v1/images/generations when a Codex OAuth image route forwards a GIF request to the upstream server and the upstream account or route is not entitled for animated GIF output (confirmed by a CLIProxyAPI collaborator in router-for-me/CLIProxyAPI#7269 on 2026-11-05). This is your case if static gpt-image-2.6-flare image generation works, but the same Codex OAuth setup fails when output_format is gif.

1. Stop retrying the same Codex OAuth generation request unchanged.
2. If a static image is acceptable, remove the GIF-only fields from the JSON body of your existing generation request:

   {
     "output_format": "gif",
     "frame_count": 4,
     "frame_duration_ms": 120
   }

3. If animated GIF output is required, move the GIF request to an API-key image path whose account and selected model have GIF preview access. The OpenAI Images API reference cited in the reports says GIF output requires preview access for the selected model, and a reporter separately observed an API-key Images API path generating a two-frame GIF with Sunburst (reported in kiss-kedaya/sub2api#195).
4. After changing access paths, review the model and image parameters in the GPT Image model guide before sending production traffic.

How to confirm it worked: The static fallback should return an image instead of the quoted HTTP 400. If you switched to an API-key image path with GIF preview access, the GIF request should return an animated image response rather than GIF output is not enabled for this request (user workaround reported in kiss-kedaya/sub2api#195).

If it still fails: Collect the request route, model name, output_format, frame_count, frame_duration_ms, proxy version, and whether the same request fails directly against the Codex image upstream. Report those details to the proxy project issue tracker; CLIProxyAPI#7269 was resolved after the reporter included the raw HTTP request, Docker image version, Codex OAuth type, and direct-upstream comparison.

GIF output is not enabled for this request on edits

HTTP 400
GIF output is not enabled for this request.

GPT Image 2.6 shows the same GIF output is not enabled for this request error on POST /v1/images/edits when a Codex OAuth edits request asks for GIF output and the upstream route rejects GIF before returning any image (reported in kiss-kedaya/sub2api#195 on 2026-11-05). This is your case if a reference-image request such as Sunburst image-to-GIF fails, while static image generation on the same OAuth group succeeds.

1. Confirm that the failing request is an edits request using the Codex OAuth native image path. In the reported sub2api case, the gateway log recorded the upstream endpoint as:

   /backend-api/codex/images/edits

2. Do not treat this as a local missing-field problem if your request already includes the GIF fields. The collaborator investigation in sub2api#195 found that the gateway passed output_format, frame_count, and frame_duration_ms through to the Codex OAuth native paths and did not perform local GIF validation.
3. If a non-animated output is acceptable, remove the GIF-only form fields from the existing edits request:

   output_format=gif
   frame_count=2
   frame_duration_ms=200

4. If animated reference-image output is required, use an API-key Images API path whose account and model have GIF preview access. The same sub2api#195 report says an official Images API plus API-key comparison could generate a two-frame Sunburst GIF from a reference image, while the Codex OAuth route returned the quoted 400.

How to confirm it worked: The static fallback should return an edited image. The API-key workaround should return a GIF response instead of failing before generation with GIF output is not enabled for this request.

If it still fails: Keep the original cURL or raw request, the upstream endpoint shown in logs, the model name, the OAuth group or access path, and the exact HTTP status. Use the general error-handling guide to structure the status, request, retry, and escalation details before filing a new issue.

How We Verified

• Generation requests: a CLIProxyAPI collaborator confirmed on 2026-11-05 that CLIProxyAPI forwards output_format: "gif", frame_count, and frame_duration_ms directly to https://chatgpt.com/backend-api/codex/images/generations and that the HTTP 400 is enforced by the upstream server because animated GIF generation is gated behind preview or feature entitlement and is not currently enabled on the Codex OAuth route (router-for-me/CLIProxyAPI#7269).
• Edit requests: a sub2api collaborator bot reported on 2026-11-05 that the gateway passes GIF fields through on /backend-api/codex/images/{generations,edits}, does not perform local GIF validation, does not retry 400 responses through a bridge fallback, and has no upstream capability signal or account-plan check for GIF support (kiss-kedaya/sub2api#195).
• API-key workaround: a user report on 2026-11-05 said the official Images API with an API key could generate a two-frame Sunburst GIF from a reference image, while the Codex OAuth native route returned the quoted 400 (kiss-kedaya/sub2api#195). Because this is a user workaround, confirm GIF preview access on the target account before relying on it.

Docs: Use the general error-handling guide to organize the status, request details, retry behavior, and escalation notes before filing a new issue. → general error-handling guide

Related Docs

• GPT Image model guide
• General API error-handling guide

Sources

• router-for-me/CLIProxyAPI#7269: GPT Image 2.6 Flare GIF output returns HTTP 400 with Codex OAuth on v8.0.56
• kiss-kedaya/sub2api#195: GPT Image 2.6 Codex OAuth GIF output returns HTTP 400
• OpenAI Images API reference cited by the issue reports