Large Discrepancy in Output Token Size Between Two Identical GPT-4o-mini Batch Runs

[_j]

You have seen the evidence of “resolving the billing issues” in your latest calls. You are billed the expected price now. There was just a bug or oversight for a short while coinciding with the release of Responses, that was fixed after a few weeks.

@_j

OpenAI also “fixed” the temporary cheap billing bug of gpt-4o-mini I reported earlier. It is back to the input token cost of images being multiplied by 33.3x. Don’t use this model if you wish to save money on image tasks.

Model	Base tokens	Tile tokens
4o, 4.1, 4.5	85	170
4o-mini	2833	5667

Unexpected token length for vision

The input token consumption by gpt-4o-mini is to ensure that there is no “cheap vision” AI model. In fact, gpt-4o-mini costs twice as much for image input as simply gpt-4o. The input token cost of an image is multiplied by 33.33x on gpt-4o-mini. 85 tokens of a “detail:low” image sent to gpt-4o becomes 2833 billed by gpt-4o-mini. Then your 6 tile + base image you are sending at detail:high, with 13x the tile and token cost of detail:low becomes the billed token cost you see. It is not more ex…