gpt-image-2-5
GPT Image 2.5
Generate the requested images through the user's ChatGPT subscription and save each prompt, references, and artifact evidence. This skill uses no API key and does not grant image-generation access. Its name does not prove which image model served a request.
For multiple images, throughput tuning, or rate limits, read batch generation before dispatching requests. It covers queue preparation, measured concurrency, retries, and collection review.
1. Choose the route and check the model requirement
Prefer the current host's subscription-backed image tool. Use the bundled Codex CLI runner only when that capability is unavailable. Apply the same prompt preparation, artifact inspection, and reporting steps to both routes.
When the user asks for an exact model, snapshot, or a guarantee of 2.5, read model selection before generating. Check the actual tool schema and installed CLI capabilities. The examined subscription routes expose no image-model selector. An image-model name in the prompt, the skill name, or codex --model does not select the image model.
If exact selection is required and the route cannot provide it, explain the limitation and stop the dependent generation. Continue any independent preparation. Do not generate an unverified substitute or automatically switch to the separately billed API. The reference records the official API identifiers for a user who chooses that route separately.
The CLI fallback requires python3, codex on PATH, and a ChatGPT login with image-generation access. The native route does not require a nested CLI login check.