- This event has passed.
Token Economics: The Story Beyond $ per Million Tokens
Token Economics: The Story Beyond $ per Million Tokens
Date and Time:
Monday, 5th October 2026
9:00 AM London Time (BST)
10:00 AM Central European Time (CEST)
7:00 PM Melbourne Time (AEDT)
Location:
Live Online Webinar/MS Teams
Hosted by:
Hume Institute for Postgraduate Studies
Centre for Artificial Intelligence in Society (AISOC)
Speaker:
Senthoor Punniamoorthy
Principal Architect, Telstra Chief Architect’s Office
Senthoor Punniamoorthy is a Principal Architect in Telstra’s Chief Architect’s Office. His work sits at the intersection of emerging technology and business strategy, translating organisational objectives into architecture and technology roadmaps that engineering teams can implement.
He has a particular interest in complex problems that do not fit neatly within a single discipline, bringing together technological, commercial and strategic considerations to identify the relevant constraints and shape them into practical solutions.
Overview:
Dollars per million tokens has become the default currency for comparing artificial intelligence models. Yet, taken in isolation, token price can be one of the least informative measures for understanding the true economics of deploying an AI system.
A headline token price tells us the cost of a unit of model input or output, but says little about how quickly that output is produced, how much context a model can process, how many tokens are required to complete a task, or the infrastructure, commercial and regulatory conditions that determine where and how a model can be deployed.
This public lecture will look beyond the headline “$ per million tokens” figure to examine what organisations are actually buying when they purchase AI model capacity. Drawing on live market benchmarks and interactive demonstrations, the session will explore the engineering trade-offs that influence token pricing and the external constraints that providers must operate within.
The lecture will consider factors including model performance, latency, context windows, token consumption, infrastructure requirements and deployment limitations, demonstrating why superficially cheaper models may not necessarily represent the lowest-cost solution for a particular task or organisational objective.
The session will ultimately propose a more rigorous framework for evaluating AI models based on the cost of achieving an outcome rather than simply the cost of an individual token. This approach provides a more meaningful basis for organisations seeking to compare models and make informed architectural, commercial and strategic decisions about artificial intelligence.
Click below for more information and to download the poster: