Pricing for GPT-5.6 seemed to be missing from the public pricing pages and retail API, while cost tools showed $0. Charges later appeared for earlier usage. Was pricing published somewhere else, and how can the charges be checked?
Did GPT-5.6 charges show up before pricing was available?
Asked by
p.62!oddity
Answers
There are conflicting reports about where the pricing was available. One account says the model announcement included rates, though not cached token pricing. Another says there was no entry in the retail API until charges appeared.
It would be worth checking both the announcement and the current pricing information, including whether cached input tokens have a separate rate. A missing retail API entry does not necessarily mean no price was published anywhere.
Reply to this answer
Check the usage dates behind the new charges, not just the day they appeared. One report says costs were added for usage going back to the model's release, even though earlier cost views had shown zero.
Compare the billed amounts with your own request and token records if you have them. If the numbers don't line up, contact support and include the affected dates, model, usage records and cost details. There are also reports of unexpectedly high amounts, but that alone doesn't establish what caused them.
Reply to this answer
Watch for revisions to the cost figures too. One account describes a dashboard amount rising from under $3 to around $15 despite its usage graphs showing no additional activity. That could indicate a reporting or billing issue, but the available information doesn't explain why it changed.
Reply to this answer
Add your answer
Answers are reviewed before publication.
5p+drainkey_294:
A zero in the cost view and a token count in the usage view can coexist, so keep both when documenting the discrepancy.