an AI model priced wrong?

Here’s gpt-realtime-2.1 on its published schedule. Text input is $4 per million tokens. Audio input is $32. Same model, same request, and the audio side costs eight times more.

My engine used to price audio tokens at the text rate. The counts were right and the rate was wrong, so anyone running voice saw a tidy number that was too low.

(Another one for my list of things I learned by building this: I’d assumed speech was billed by the minute. It’s turned into audio tokens, and those have their own price.)

If you’re estimating a voice feature, here’s what the gap looks like. My calculator uses a rough 600 audio tokens per minute of speech. Three minutes of someone talking is about 1,800 tokens. At the audio rate that’s $0.0576. At the text rate it’s $0.0072. Over 2,000 sessions a month, $115.20 against $14.40, on the input side alone.

There’s no rule of thumb to fall back on either. On gemini-3.1-flash-lite, audio input is $0.50 against $0.25 for text, so twice the price. A multiplier you copy from one model is wrong on the next.

The awkward case is a model that publishes no audio rate at all. llm-track shows that rate as unknown and marks the estimate incomplete. A visible gap gives you something to chase. A made-up rate gives you a confident number nobody questions.

The engine now prices text in, text out, audio in and audio out as separate components, each at its own published rate.

So, the answer: inside one model, audio and text are billed as different things. Price them separately, or your voice estimate starts out short.