May 30, 07:47 PM

Ollama Adds Thinking Feature With Multiple Reflection Phases to API, CLI, Libraries, Including DeepSeek-R1-0528 Model

Ollama has introduced a new feature that enables its AI models to engage in distinct thinking processes separate from their responses. This enhancement is particularly beneficial for thoughtful models such as DeepSeek-R1-0528, allowing for multiple periods of reflection within the reasoning chain. Users can also disable the thinking feature to receive direct responses when preferred. The thinking capability is integrated across Ollama's API, command-line interface (CLI), and libraries, providing developers with flexible tools to leverage this functionality. Early observations highlight improvements in the model's ability to avoid repetitive search loops and exhibit clearer phases of thought during processing.

#Ollama #DeepSeek

Written with ChatGPT (GPT-4).

Sources

TestingCatalog News 🗞@testingcatalog
1 year ago
ollama now supports thinking traces 🤖 This feature can be tested with the newest DeepSeek-R1-0528 model https://t.co/Yutj5v40GJ https://t.co/waFbJjcrT6
Jeffrey Morgan@jmorgan
1 year ago
Ollama now has thinking built into the API, CLI and libraries!! https://t.co/yJRX5otvn8
ollama@ollama
1 year ago
Ollama can now think! 🤔🤔🤔 For thinking models, and especially useful for very thoughtful models like DeepSeek-R1-0528, Ollama can separate the thoughts and the response. Thinking can also be disabled! This is useful for getting a direct response. This works across https://t.co/kIInaJ5aMp

Ollama Adds Thinking Feature With Multiple Reflection Phases to API, CLI, Libraries, Including DeepSeek-R1-0528 Model

Sources

Additional media

Similar Stories