Logit_bias for gpt-4o-audio-preview

https://platform.openai.com/docs/api-reference/chat/create#logit_bias

To Reproduce

When I pass logit_bias into the text-only models I get substantially different results than when I pass it into gpt-4o-audio-preview. It doesn’t reject the logit_bias which makes me believe it is valid, but I’m unusure it’s having an effect. Could you confirm? Thanks!

Also, will you be releasing log_probs for audio anytime soon? Thanks!