Ollama

DeepEval allows you to use any model served by Ollama to run evals, either through the CLI or directly in Python. Some capabilities, such as multimodal support, are detected from a known-model list.

Environment Setup

DeepEval can use a local Ollama server (default: http://localhost:11434). Optionally set a custom host:

# .env.local
LOCAL_MODEL_BASE_URL=http://localhost:11434

Command Line

To configure your Ollama model through the CLI, run the following command. Replace deepseek-r1:1.5b with any Ollama-supported model of your choice.

deepeval set-ollama --model=deepseek-r1:1.5b

You may also specify the base URL of your local Ollama model instance if you've defined a custom port. By default, the base URL is set to http://localhost:11434.

deepeval set-ollama --model=deepseek-r1:1.5b \
    --base-url="http://localhost:11434"

Python

Alternatively, you can specify your model directly in code using OllamaModel from DeepEval's model collection.

from deepeval.models import OllamaModel
from deepeval.metrics import AnswerRelevancyMetric

model = OllamaModel(
    model="deepseek-r1:1.5b",
    base_url="http://localhost:11434",
    temperature=0
)

answer_relevancy = AnswerRelevancyMetric(model=model)

from deepeval.metrics import AnswerRelevancyMetric

answer_relevancy = AnswerRelevancyMetric(
    model="deepseek-r1:1.5b",
)

There is ONE mandatory parameter and THREE optional parameters when creating an OllamaModel:

• [Optional] model: A string specifying the name of the Ollama model to use. Defaults to OLLAMA_MODEL_NAME if not passed; raises an error at runtime if unset.
• [Optional] base_url: A string specifying the base URL of the Ollama server. Defaults to LOCAL_MODEL_BASE_URL if not passed; falls back to http://localhost:11434 if unset.
• [Optional] temperature: A float specifying the model temperature. Defaults to TEMPERATURE if not passed; falls back to 0.0 if unset.
• [Optional] generation_kwargs: A dictionary of additional generation parameters forwarded to Ollama’s chat(..., options={...}) call.

Available Ollama Models

Below is a list of commonly used Ollama models:

• deepseek-r1
• llama3.1
• gemma
• qwen
• mistral
• codellama
• phi3
• tinyllama
• starcoder2