The AI Token Price Index (TPI) project has begun tracking the average blended cost per million tokens for leading AI models, capturing market dynamics and industry pricing cycles.
What Happened
The AI Token Price Index (TPI) tool has been launched, providing data on the cost of using advanced language models. The current index includes 19 participants. Notably, access to the Fable 5 model was temporarily restricted on June 12, 2026, due to US export controls.
Context
The index tracks the blended cost metric, which allows for the recording of market cycles: price increases upon the release of new SOTA models and their subsequent decline as cheaper alternatives and technological optimizations emerge. This moves token costs from a category of opaque data into the category of market benchmarks.
Why It Matters for the Industry
For the industry, TPI creates a transparent standard for evaluating computation costs. This allows companies to effectively plan budgets for LLM API usage, develop automated cost-aware routing tools, and standardize approaches to calculating the unit economics of AI services.
Why It Matters for Users
For users and developers, this provides the opportunity to understand the real economics of AI and conduct accurate comparative cost analyses across different model categories (Frontier, Mid, Fast) when designing RAG architectures or agentic systems.
Sources
Author
Look at AI, Editorial Team