[Responses API] GPT 5 ignores the detail parameter on image inputs

[_j]

@manouchehri

OpenAI for GPT 5.2.

GPT-5.2 uses “patches” algorithm for vision, not “tiles”. That type of vision input does not have a low detail.

Of course noted nowhere and the pricing not in calculators - because the price of those images on gpt-5.2 with no detail low option is also about 3x more that GPT-5.1:high.

API AI computer vision pricing - models compared visually

Vision - input image pricing I though it would be more useful to show you a comparison of what these models cost to run with input images, rather than you getting some abstract formulas for tokens and multipliers and costs per million. By model; 1920x1080, vs 768x512 [image] [image] Notably: OpenAI is multiplying the token cost of GPT-5.2 by 1.2x – mentioned NOWHERE, the most expensive model since original GPT-4o. High-cost models (vs gpt-4o) [image] The large image will be downsized whet…