The unit of text an LLM works with, which may be a word, part of a word or punctuation. Billing, limits and context windows are all counted in these units.
Read more: Microsoft Learn
In the Ultra Transcenders books
Each book explains Token in context, with comparison tables and the common traps.
Terms in this definition
- LLM
Large language model, usually a transformer network with billions of parameters that learnt to predict the next token from vast amounts of text. Compared with a small language model it is more capable across tasks, but slower and costlier.
- ALL
A DAX function that ignores any filters and gives back every row of a table or every value of the named columns. Used within CALCULATE, it works as a modifier that clears filters, although REMOVEFILTERS states that intent more clearly where it is available.
Related terms
- Account SAS
Shared access signature created with an account key. One token may cover multiple services (ss), resource types (srt) and permissions (sp), service-level operations included, but turning off Shared Key authorisation blocks it.
- Agent Monitoring Dashboard
View in Microsoft Foundry, powered by Application Insights, that tracks token consumption, latency, successful runs and evaluation scores and can raise alerts. It gives visibility at runtime and is not a gate for releases.
- AI gateway in Azure API Management
Collection of API Management gateway capabilities, offered in every tier rather than as its own product, for placing model APIs, MCP servers and A2A agent APIs behind caller authentication, token metering and content screening.
- API Management llm-token-limit policy
Inbound AI gateway policy that caps token usage per counter key. Exceeding tokens-per-minute produces 429, an exhausted token quota produces 403, and estimate-prompt-tokens lets it reject before calling the backend.
- API Management validate-jwt policy
API Management inbound policy validating a JWT's issuer, audience, signature and required claims, for instance using the Microsoft Entra OpenID configuration. By default it answers 401 when a token is absent or invalid.
- APIM subscription key
Caller key supplied either as a query-string parameter or in the Ocp-Apim-Subscription-Key header. API Management's gateway checks it as an APIM subscription key, which is not a Microsoft Entra token, while Foundry Tools like Content Safety read the resource key from it.
- App roles
Roles declared by an API in its app registration. When one app calls another using client credentials, the assigned roles appear in the token's
rolesclaim. - Attention
In a transformer, the mechanism determining the weight each other token in the context carries when predicting the next one.