Cost estimates for traces

Hi everyone,

When we look at the pricing estimates in LangSmith for our traces, they look off. For example, a conversation with multiple traces that has ~700k input tokens has a cost estimate of ~$0.30. But almost all the models used are gpt-4.1, and it costs $2/1M tokens, so we should see at least a $1 or so for this conversation.

Are we missing something?

Thanks,
Eugene

Hi Eugene, could you share a public trace with us where this was problematic?