Hi Roshan J. we currently only track token count if the APIs provide it. We did this so the tracing has no impact on your runtime. We will have a mapping of all the tokenizers in the future and do the calculations for you post-collection. Hope that helps.