Glossary
Token
The unit models read and bill in, roughly three-quarters of an English word, and more for Indian scripts.
Text is split into tokens before a model processes it. Pricing, context limits and latency are all measured in tokens.
Indian-language text tokenises less efficiently than English, so the same sentence in Hindi or Tamil can cost noticeably more. That matters when budgeting multilingual systems.
Practically, this changes multilingual budgets. A support assistant handling English, Hindi and Tamil will not cost three equal thirds, the Indic-script conversations can cost noticeably more per exchange for the same amount of meaning. Budget from measured samples of real conversations rather than from an English baseline.
There is a second-order effect worth planning for. Because Indic scripts consume more tokens per unit of meaning, they also consume more of the context window, so a document set that fits comfortably in English may not fit in Tamil or Hindi. Chunking strategy has to account for the language, not just the format.
Two practical habits keep costs predictable: measure tokens per request in development rather than discovering them in production, and set a hard ceiling per feature. Runaway loops are the single most common cause of an unexpected bill, and a ceiling turns that from a month-end surprise into an alert.
Related terms, in context
The concepts you almost always meet alongside token.
- Context window
- How much text a model can consider at once, measured in tokens.
- Inference cost optimisation
- Reducing what AI features cost to run, usually 40 to 70% without measurable quality loss.
Where this shows up in our work
Token is not an abstraction for us. It is a decision we make on live projects. It shows up most directly in llm cost optimisation, where getting it wrong has a cost someone can measure.
If you are evaluating a vendor on this, the useful question is not whether they can define the term. It is what they measure, what they would refuse to do, and what happens in their system when the assumption behind token stops holding.
Questions
What is Token?
The unit models read and bill in, roughly three-quarters of an English word, and more for Indian scripts.
Does Orqent Labs build this?
Yes, LLM Cost Optimisation. We work across India, covering all 19,238 PIN codes remotely.
Building something that involves token?
We will tell you honestly whether it is the right approach for your problem.
Or email bd@dtrasglobal.com · call +91 74118 77878