# RapidSeek

**Repository Path**: RapidAI/RapidSeek

## Basic Information

- **Project Name**: RapidSeek
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: main
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2025-08-01
- **Last Updated**: 2025-08-01

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# RapidSeek
An Economical Large Language Model Inference Cluster System for DeepSeek MOE Models (such as DeepSeek V3 and R1)


# Featurs:

- C++ and Python-based 
- efficient kernels in C++ 
- for multiple machine clusters with a cost below 1,5000 dollars per machine
- Achieve 20 tokens per second per user.

## Models supported
* DeepSeek-V2-Lite 

* DeepSeekV3

* DeepSeekR1

### Supported  GPUs
- Nvidia
- AMD
- Intel Arc
  
### Recommended hardware


### Special thanks to 

- llamafile https://github.com/Mozilla-Ocho/llamafile
- Marlin  https://github.com/IST-DASLab/marlin
- MLA  https://github.com/ikawrakow/ik_llama.cpp