{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreif2yyuj7bacyjlkxsmwhv5cejoay3devm2my4bzf3ivclaubcuv64",
    "uri": "at://did:plc:lk3jfj3zq4k4wxnk474axylu/app.bsky.feed.post/3mnomec3wgdw2"
  },
  "path": "/t/openai-must-document-the-input-image-pricing-of-gpt-image-2/1382940#post_1",
  "publishedAt": "2026-06-07T06:15:20.000Z",
  "site": "https://community.openai.com",
  "textContent": "Absolutely no documentation is provided that describes the billing received for input images of various sizes on the edits endpoint, when using gpt-image-2\n\nAdditionally, nothing describes the context and typical or maximum received image count from use of the image editing tool on the Responses API endpoint, which obviously runs a parallel context consuming model with double-billing and no usage report.\n\nThe costs billed in tokens do not relate in any way to the “patches” algorithm of gpt-image-1, where images are described as downscaled to 512px for the shortest dimension, and then the only thing that could increase would be the longest dimension, by 512px steps. Also, input_fidelity is not an API option accepted but described as “always on”, for a described additional 4k or 6k billed, whereas this is forced “high” on gpt-image-1.5, ignoring “low”, contrary to documentation in multiple places.\n\nInstead, increasing the width, there are steps that are not monotonic:\n\nwidth x height → image_tokens\n511 x 512 → 1024\n512 x 512 → 1024 delta=+0\n513 x 512 → 1024 delta=+0\n1023 x 512 → 512 delta=-512\n1024 x 512 → 512 delta=+0\n1025 x 512 → 528 delta=+16\n1535 x 512 → 768 delta=+240\n1536 x 512 → 768 delta=+0\n1537 x 512 → 833 delta=+65\n2047 x 512 → 1408 delta=+575\n2048 x 512 → 1408 delta=+0\n2049 x 512 → 1430 delta=+22\n\nThen take increasing a square image, where it responds to smaller sizes, but then has a stalling point:\n\nwidth x height → image_tokens\n256 x 256 → 256\n384 x 384 → 576 delta=+320\n768 x 768 → 1024 delta=+448\n1024 x 1024 → 1024 delta=+0\n1536 x 1536 → 1521 delta=+497\n\nOr then a 2:1 aspect ratio, where we initially suspect a pattern similar to before, but then that breaks down.\n\nwidth x height → image_tokens\n512 x 256 → 512\n1536 x 768 → 1152 delta=+640\n2048 x 1024 → 1458 delta=+306\n4096 x 2048 → 1458 delta=+0\n\nRequests were made at an output size of “1440x480”, quality:low.\nA ratio of 1521 tokens billed for input image to 54 tokens billed for output image.\nIncreasing the quality did not change the costs, nor did output size.\n\n## Publish the algorithm and its steps for normalizing, resizing, and billing by the resulting input image.\n\n## Document clearly what the AI is “seeing” in terms of billed semantic patches.\n\n100% accurate cost estimation of a request is required.",
  "title": "OpenAI *must* document the input image pricing of gpt-image-2"
}