cortexso/deepseek-r1

Overview

DeepSeek developed and released the DeepSeek-R1 series, featuring multiple model sizes fine-tuned for high-performance text generation. These models are optimized for dialogue, reasoning, and information-seeking tasks, providing a balance of efficiency and accuracy while maintaining a smaller footprint compared to their original counterparts.

The DeepSeek-R1 models include distilled and full-scale variants of both Qwen and Llama architectures, catering to various applications such as customer support, conversational AI, research, and enterprise automation.

Variants

DeepSeek-R1

No	Variant	Branch	Cortex CLI command
1	DeepSeek-R1-Distill-Qwen-1.5B	1.5b	cortex run [WIP]
2	DeepSeek-R1-Distill-Qwen-7B	7b	cortex run [WIP]
3	DeepSeek-R1-Distill-Llama-8B	8b	cortex run [WIP]
4	DeepSeek-R1-Distill-Qwen-14B	14b	cortex run [WIP]
5	DeepSeek-R1-Distill-Qwen-32B	32b	cortex run [WIP]
6	DeepSeek-R1-Distill-Llama-70B	70b	cortex run [WIP]

Each branch contains a default quantized version:

• Qwen-1.5B: q4-km
• Qwen-7B: q4-km
• Llama-8B: q4-km
• Qwen-14B: q4-km
• Qwen-32B: q4-km
• Llama-70B: q4-km

Use it with Jan (UI)

1. Install Jan using Quickstart
2. Use in Jan model Hub:

   cortexso/deepseek-r1 [WIP]
   cortexso/deepseek-r1 [WIP]
   

Use it with Cortex (CLI)

1. Install Cortex using Quickstart
2. Run the model with command:

   cortex run [WIP]
   

   or

   cortex run [WIP]
   

Credits

• Author: DeepSeek
• Converter: Homebrew
• Original License: License
• Papers: DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning