162 5 months ago

DeepSeek's first-generation of reasoning models with comparable performance to OpenAI-o1, including six dense models distilled from DeepSeek-R1 based on Llama and Qwen.

thinking 671b

Models

View all →

Readme

Note: this model requires Ollama 0.9 or later.

Before running the instruction, please set the num_thread to half of your current CPU thread count, otherwise it may slow down your computer. Here is an example:

/set parameter num_thread 32