Understand Genesys Agent Copilot AI models and LLM input

Genesys Agent Copilot uses a combination of Genesys native and AWS Bedrock models, depending on the feature, use case, and dialect.

Note: Large language model (LLM) tokens are not the same as Genesys Cloud AI Experience tokens. For more information about Genesys Cloud AI Experience tokens and pricing, see Genesys Cloud AI Experience tokens metering and pricing.

Genesys Agent Copilot AI models and LLM input

Genesys Agent Copilot uses a combination of Genesys native and AWS Bedrock models, depending on the feature, use case, and dialect.


Language Region Language
tag		 Knowledge
surfacing		 Knowledge
surfacing answer
highlight		 Intent-based
NBA 		 Summarization Advanced summarization Wrap-up code prediction
English Australia  en-AU Genesys Native

AWS Bedrock

 Genesys Native

Token input
< 6000: Genesys Native

Token input
> 6000: AWS Bedrock

Token input
< 6000: Genesys Native

Token input
> 6000: AWS Bedrock

Token input
< 6000: Genesys Native

Token input
> 6000: AWS Bedrock
Great Britain  en-GB Genesys Native AWS Bedrock Genesys Native

Token input
< 6000: Genesys Native

Token input
> 6000: AWS Bedrock

Token input
< 6000: Genesys Native

Token input
> 6000: AWS Bedrock

Token input
< 6000: Genesys Native

Token input
> 6000: AWS Bedrock
United States en-US Genesys Native AWS Bedrock Genesys Native

Token input
< 6000: Genesys Native

Token input
> 6000: AWS Bedrock

Token input
< 6000: Genesys Native

Token input
> 6000: AWS Bedrock

Token input
< 6000: Genesys Native

Token input
> 6000: AWS Bedrock
Dutch Netherlands  nl-NL

Genesys Native

 AWS Bedrock* Genesys Native AWS Bedrock AWS Bedrock* AWS Bedrock
French Canada  fr-CA Genesys Native AWS Bedrock* Genesys Native AWS Bedrock AWS Bedrock* AWS Bedrock*
France fr-FR Genesys Native AWS Bedrock* Genesys Native AWS Bedrock AWS Bedrock* AWS Bedrock*
German Germany  de-DE Genesys Native AWS Bedrock* Genesys Native AWS Bedrock AWS Bedrock* AWS Bedrock*
Italian Italy it-IT Genesys Native AWS Bedrock* Genesys Native AWS Bedrock AWS Bedrock* AWS Bedrock*
Japanese Japan jp-JP Genesys Native AWS Bedrock* Genesys Native AWS Bedrock AWS Bedrock* AWS Bedrock*
Portuguese Brazil pt-BR Genesys Native AWS Bedrock* Genesys Native AWS Bedrock AWS Bedrock* AWS Bedrock*
Portugal pt-PT Genesys Native AWS Bedrock* Genesys Native AWS Bedrock AWS Bedrock* AWS Bedrock*
Spanish United States es-US Genesys Native AWS Bedrock* Genesys Native AWS Bedrock AWS Bedrock* AWS Bedrock*
Spain es-ES Genesys Native AWS Bedrock* Genesys Native AWS Bedrock AWS Bedrock* AWS Bedrock*

* Roadmap

Tokens overview

Tokens represent segments of text, like parts of words, spaces, or punctuation. The API processes input and breaks that input down into tokens, which do not always align with whole words. Some key approximations include:

  • One token equals approximately four characters in English
  • One token equals approximately ¾ of a word
  • 100 tokens equal approximately 75 words

 Or: 

  • 1–2 sentences equal approximately 30 tokens
  • One paragraph equals approximately 100 tokens
  • 1,500 words equal approximately 2048 tokens

For example, a popular saying like “Carpe diem” might have 10 tokens. A company’s vision statement could be about 500 tokens. A classic poem, such as “The Road Not Taken” might contain around 1,700 tokens.

Language differences

Tokenization varies by language. For instance, “Bonjour tout le monde,” which means “Hello everyone” in French, contains five tokens, despite being just 19 characters. Non-English texts may require more tokens, which impact API costs.

How tokens generate

The API converts words into tokens based on their context, processes them, and then translates the tokens back into readable text. The same word may split into different tokens, depending on its placement. For instance, the word “blue” could have a different token if it appears with a trailing space or starts with a capital letter.

Examples:

  • The token for “blue” with a trailing space might be “4321.”
  • The token for “Blue” (capitalized with a space) could be “4521.”
  • The token for “Blue” (capitalized at the start) might be “5000.”

Common tokens have lower values. Punctuation, like commas, tends to have a consistent token ID due to its frequent use. Words may have different token IDs based on their case or position in a sentence.