M

Minicpm 2B Sft Fp32

Developed by openbmb
MiniCPM is a series of edge-side large language models jointly open-sourced by Mianbi Intelligence and the Natural Language Processing Laboratory of Tsinghua University. The main language model, MiniCPM-2B, has only 2.4 billion non-word embedding parameters.
Downloads 218
Release Time : 1/30/2024

Model Overview

MiniCPM is an efficient large language model for edge devices. After supervised fine-tuning (SFT) and direct preference optimization (DPO), it performs excellently in multiple evaluations, supports both Chinese and English, and can be deployed on mobile devices.

Model Features

Efficient edge-side deployment
After Int4 quantization, it can be deployed and inferred on mobile phones, and the streaming output speed is slightly higher than the human speaking speed.
Excellent performance
In public evaluations, its performance is similar to that of Mistral-7B, and its overall performance surpasses models such as Llama2-13B, MPT-30B, and Falcon-40B.
Multimodal support
The edge-side multimodal large model MiniCPM-V built based on MiniCPM-2B outperforms models of the same scale.
Low-cost development
Only one 1080/2080 graphics card is needed to efficiently perform parameter-efficient fine-tuning, and the secondary development cost is relatively low.

Model Capabilities

Text generation
Dialogue system
Multimodal understanding
Code generation
Mathematical reasoning

Use Cases

Intelligent assistant
Question-answering system
Answer various questions raised by users
Performs excellently in Chinese, mathematics, and code capabilities
Education
Learning assistance
Help students answer learning questions
Performs well in mathematical reasoning ability
Development assistance
Code generation
Assist developers in generating code snippets
Performs excellently in code ability evaluations
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
© 2025AIbase