#general
Oct 9October 13, 2026
Hey guys! I already reported some issue on OpenRouter pricing miscalculations, but I noticed another one for embeddings, when routed via OpenRouter. It consistently computes the embedding price at $50/input Mtok, despite the corresponding model entry in the catalog being correctly set to $0.13/Mtok for text-embedding-3-large. Ultimately, I added an OpenAI API Key, and now the calculation is correct, but the previous recorded costs remain erroneous.
Is this a known issue on your side? And is there a way to recompute the past entries?
:archestra-love:1
5 repliesRead-only live mirror of Archestra.AI Slack
👋Join the discussion withAI enthusiasts!Hey guys! I already reported some issue on OpenRouter pricing miscalculations, but I noticed another one for embeddings, when routed via OpenRouter. It consistently computes the embedding price at $50/input Mtok, despite the corresponding model entry in the catalog being correctly set to $0.13/Mtok for text-embedding-3-large. Ultimately, I added an OpenAI API Key, and now the calculation is correct, but the previous recorded costs remain erroneous.
Is this a known issue on your side? And is there a way to recompute the past entries?
:archestra-love:1
5 replies
hello! in fact a fix will go out for this in v1.4.0-rc.38, which should be released today/tomorrow 🙂
Yayy thanks!
Will there be a way to fix existing cost logs?
hmm not currently! are you relying on cost reporting from those past logs for something?
(this is a bit of a trickier thing to implement - ex. exactly which logs to update, etc)
you could ask your coding agent nicely to do a one-off database update to surgically update only the LLM logs you care about 😁 (but no in app way of doing this currently)