Neura Market
MOSS logo

MOSS

Free

An open-source tool-augmented conversational language model from Fudan University

Writing AssistantsFreeFree tier
Inputs: text, codeOutputs: text, code
Type
Open Source
Company
Fudan University
Visit Website GitHub (12.1k stars)

About MOSS

MOSS is an open-source conversational language model developed by Fudan University, designed to resemble ChatGPT's capabilities. It can follow user instructions to perform a variety of natural language tasks including question answering, text generation, summarization, and code generation. MOSS is also trained to challenge incorrect premises and reject inappropriate requests, adhering to the principles of being helpful, honest, and harmless (HHH). During the research preview, usage is free, and the model is open-sourced to facilitate future research. MOSS learns by interacting with humans and other AI models, but its relatively small parameter count (compared to ChatGPT) and limited training data result in certain limitations, such as poorer performance on non-English languages and occasional failures in instruction following.

Key Features

Open-source and free to use during research preview
Tool-augmented conversational abilities
Capable of question answering, text generation, summarization, and code generation
Can challenge incorrect premises and reject inappropriate requests
Designed to be helpful, honest, and harmless (HHH)
Learns from conversations with humans and other AI models

Pros & Cons

Pros
  • Open-source, allowing full transparency and customization
  • Free to use during the research preview
  • Adheres to HHH principles (helpful, honest, harmless)
  • Can challenge incorrect premises and reject inappropriate requests
  • Supports multiple natural language tasks including code generation
  • Actively improved based on user feedback
Cons
  • Lags far behind ChatGPT due to limited data, compute, and model capacity
  • Performs poorly on languages other than English (Chinese support in progress)
  • Contains limited world knowledge, leading to potentially misleading or false information
  • Sometimes fails to follow instructions or requires multiple regeneration attempts
  • Can be prompted to generate unethical or harmful responses
  • Access is currently limited to invited users due to compute constraints

Best For

Question answering and knowledge retrievalText generation and creative writingSummarization of long documentsCode generation and programming assistanceConversational AI and chatbot developmentResearch and experimentation in language models

Alternatives to MOSS

CDial-GPT

A Large-scale Chinese Short-Text Conversation Dataset and Chinese pre-training dialog models

ChatLM-mini-Chinese

中文对话0.2B小模型（ChatLM-Chinese-0.2B），开源所有数据集来源、数据清洗、tokenizer训练、模型预训练、SFT指令微调、RLHF优化等流程的全部代码。支持下游任务sft微调，给出三元组信息抽取微调示例。

Chat-Haruhi-Suzumiya

Chat凉宫春日, An open sourced Role-Playing chatbot Cheng Li, Ziang Leng, and others.

DialoGPT

Large-scale pretraining for dialogue

GPT2-Chinese

Chinese version of GPT2 training code, using BERT tokenizer.

gpt-2-simple

Python package to easily retrain OpenAI's GPT-2 text-generating model on new texts

FAQ

What are the differences between MOSS and ChatGPT?
MOSS has far fewer parameters than ChatGPT. MOSS learns by talking to humans and other AI models, while ChatGPT uses Reinforcement Learning from Human Feedback (RLHF). MOSS is open-sourced to facilitate research, whereas ChatGPT may not be.
What can MOSS do?
MOSS can perform various natural language tasks including question answering, generating text, summarizing text, and generating code. It is also able to challenge incorrect premises and reject inappropriate requests.
Is MOSS free to use?
During the research preview, usage of MOSS is free. Users are invited via a waitlist to try the service and provide feedback.