Implementation:Cohere ai Cohere python SingleGeneration Model

Knowledge Sources	Cohere Python SDK
Domains	SDK, Text Generation
Last Updated	2026-02-15 14:00 GMT

Overview

SingleGeneration is a Pydantic model representing a single text generation result returned by the Cohere Generate API.

Description

The SingleGeneration model encapsulates one generation output from the Cohere text generation endpoint. Each instance contains the generated text, a unique identifier, and optional metadata including the generation index (when multiple generations are requested) and token-level log-likelihoods for model evaluation.

The token_likelihoods field is only populated when the return_likelihoods parameter is set to GENERATION or ALL in the API request. The likelihood value represents the average log-likelihood of the entire generated string, which is useful for evaluating model performance, especially with custom models.

Usage

Use SingleGeneration when working with responses from the Cohere Generate API. It is typically returned as part of a list within a Generation response object when num_generations is greater than zero.

Code Reference

Source Location

• Repository: Cohere Python SDK
• File: src/cohere/types/single_generation.py

Signature

class SingleGeneration(UncheckedBaseModel):
    id: str
    text: str
    index: typing.Optional[int] = pydantic.Field(default=None)
    likelihood: typing.Optional[float] = None
    token_likelihoods: typing.Optional[
        typing.List[SingleGenerationTokenLikelihoodsItem]
    ] = pydantic.Field(default=None)

Import

from cohere.types import SingleGeneration

I/O Contract

Fields

Field	Type	Required	Description
id	str	Yes	Unique identifier for this generation.
text	str	Yes	The generated text content.
index	Optional[int]	No	The nth generation index. Only present when num_generations is greater than zero.
likelihood	Optional[float]	No	Average log-likelihood of the entire generated string.
token_likelihoods	Optional[List[SingleGenerationTokenLikelihoodsItem]]	No	Per-token log-likelihoods. Only returned if return_likelihoods is set to GENERATION or ALL. The first token will not have a likelihood.

SingleGenerationTokenLikelihoodsItem Fields

Field	Type	Required	Description
token	str	Yes	The token string.
likelihood	float	Yes	The log-likelihood value for this token.

Usage Examples

from cohere.types import SingleGeneration

# Accessing fields from a generate response
response = client.generate(
    prompt="Write a tagline for a coffee shop",
    num_generations=3,
    return_likelihoods="GENERATION"
)

for generation in response.generations:
    print(f"ID: {generation.id}")
    print(f"Text: {generation.text}")
    print(f"Index: {generation.index}")
    print(f"Likelihood: {generation.likelihood}")

    if generation.token_likelihoods:
        for token_info in generation.token_likelihoods:
            print(f"  Token: {token_info.token}, Likelihood: {token_info.likelihood}")

Related Pages

• Environment:Cohere_ai_Cohere_python_Python_SDK_Runtime