The same 400K context, a quarter of the input price
GPT-5 Image Mini keeps GPT-5 Image's 400K context and file input while cutting the input price to a quarter. The trade is fidelity, so it fits iteration and volume rather than final output.
Try GPT-5 Image Mini