{
"$type": "site.standard.document",
"bskyPostRef": {
"cid": "bafyreie2bo5i3js7w2pf4e6epgz3u6vjc4w5tlgh2ejifyg5vdyauslkvy",
"uri": "at://did:plc:lk3jfj3zq4k4wxnk474axylu/app.bsky.feed.post/3mfw2mhcku5e2"
},
"path": "/t/fine-tuned-4o-errors-using-the-responses-api-with-image-input/1375316#post_4",
"publishedAt": "2026-02-28T09:15:25.000Z",
"site": "https://community.openai.com",
"tags": [
"Fine-tuned Model Vision Capabilities Lost",
"Bugs"
],
"textContent": "The first thing I set, knowing that it would loop, was the _maximum output_. Useless.\n\nThis fault will incur a huge bill, unstoppable by `max_output_token`s. Here, setting it to just above the minimum value accepted (which reports the same 35k input usage that would have been 301t for the small image otherwise):\n\n\n\nVarying the ranges of max_output_tokens doesn’t vary the bill of the same amount. Nor does terminating a streaming output. I imagine if you sent a message with many non-tiny images, the potential cost would not be limited at all by the model context, like you show. It probably also can re-bill for the “deny images” system message of 200+ tokens that is exempted, that OpenAI attempts to make secret.\n\nThis massively violates the API promise of max_output_tokens being a billing limit.\n\nTo stop any support time-wasting and render “provide a request id” useless to delay action:\n\n * Empty response and nonstop output:\n`x-request-id req_f48bdc55509442b6a033864f1d7aea8b`,\n`req_54a906c37e774230a84105c9b04f0caf`\n * Empty response and terminates on one:\n`x-request-id req_e2422e2c709e496582b656864bd0ddf1`\n\n\n\nThe playground seems to be only demonstrating one empty assistant output when it happens about 50% of the time currently, yet even at max_output_tokens=16 and the tiniest of images for 282t input:\n\n* * *\n\n### Chat Completions:\n\n1 in 10 or so, exactly the same fault as before there too, instead of going loopy:\n\n\n\nRequest ID for the Chat Completions API returning 400:\n`x-request-id req_eafcb2667c344ec78350e71be2405bd3`\n\nThe prior thread from two weeks ago with two weeks of inaction, then on gpt-4o, gpt-4.1 and gpt-4.1 mini fine tunes:\n\nFine-tuned Model Vision Capabilities Lost Bugs\n\n> Start requesting refunds for all the gpt-4.1 training credits in dollar figures - and it doesn’t have to be one trained on vision, because this makes any ft model unavailable for use in the image input pattern you want. Might get fixed.",
"title": "Fine-tuned 4o errors using the responses API (with image input)"
}