GPT Image in the Batch API: plan for completion, not instant previews

The update

OpenAI added Batch API support for GPT Image 1.5, chatgpt-image-latest, GPT-image-1, and GPT-image-1-mini on February 10, 2026.

Separate interactive and scheduled work

A designer waiting for a preview has a different expectation from a catalog job running ahead of a planned release. Put explicit deadlines on each workload before considering asynchronous processing.

Batch support is a provider API feature. Verify that your selected service and model support the required batch interface before building around it. A standard image endpoint and a batch submission endpoint are different integration contracts.

Design for partial completion

Give every item a stable application identifier. Track which items completed, failed, or still need review, and reconcile the results before creating a retry set. Avoid resubmitting an entire job simply because a small subset needs attention.

Store the brief and output destination with each item. When results arrive later, the application should not depend on a browser tab or an in-memory session still being open.

Compare the full economics

Consider API charges, storage, result reconciliation, and the cost of missed deadlines. Keep provider batch pricing separate from XMH.NET volume pricing; do not assume the discounts combine. For time-sensitive work, operational simplicity can matter as much as a lower unit price.