Count of input token in playground in Non English language

Hello every body,
I create a assistant with file attached. I ask short question about 27 Token but I see 8460 Input token in playground. language is none English and I know that its possible that input token in non english language increase but input token increase about x300. I check cost from usage dashboard and cost item confirm increase in input token abnormaly. why this happen?

The contents of the file also count as input tokens, assuming you are using the Assistants API.

1 Like

Thanks for your reply, in this way cost of input grows, Is there any method that dont need to use whole file as input token?

1 Like

Yes, this is a typical challenge that requires you to create a custom retrieval mechanism.
Based on the question, you provide the model with only the most relevant parts of your input file. This is called RAG, and in this context based on embeddings.

On the other hand, when using the Assistants API we get a standard solution with high token usage but it’s also very likely that the LLM can answer the question because all context is provided.

I suggest you read up on this topic and decide if it’s worth it for you and if you have more questions, feel free to ask here in the community.

You will be most interested in question answering: