MOSS

MOSS is an open-source conversational language model that supports bilingual dialogue in Chinese and English and tool use.

MOSS is an open-source conversational language model developed by Fudan University that supports Chinese and English bilingual dialogue and tool calls. The moss-moon series models feature 16 billion parameters and are pretrained on approximately 700 billion words including Chinese, English, and code. It undergoes dialogue instruction tuning, plugin enhancement learning, and human preference training to support multi-turn conversations and plugins. Supported plugins include web search, text-to-image generation, a calculator, and an equation solver. The model can run locally with varying VRAM requirements depending on FP16, INT8, or INT4 quantization levels, and supports deployment on NVIDIA GPUs like A100, A800, and 3090. It is designed for developers and researchers who need an open conversational model with tool integration capabilities.

MOSS at a glance

Best for: Software engineer, Researcher
Runs on: Linux, Windows, macOS, API, CLI, Web, self-hosted, open source
Pricing: Free open-source model weights and code; commercial use requires a signed agreement and questionnaire without fees.
Released: April 2023
Category: AI chat assistant
Website: github.com/OpenMOSS/MOSS

Key features

Support Chinese and English bilingual dialogue
Integrate search engine plugin
Generate images from text
Perform calculations
Solve equations
Provide multi-turn conversation capabilities
Run with 4bit and 8bit quantization

Use it for

Solve math application problems
Solve equations
Generate images
Conduct Chinese conversations
Perform coding tasks

Alternatives to MOSS

LM Studio — Runs open source language models locally with real-time voice transcription.
LibreChat — Unifies multiple AI models and tools in one customizable, open-source platform.
5ire — It supports local knowledge base management and Model Context Protocol servers.
llmspy.org — Self-host access to 530+ AI models across 24 providers with built-in tools.
Text Generation WebUI — Run local LLMs with multiple backends, multimodal input, and tool calling 100% offline.
Chuanhu Chat — Provides a clean user interface with multi-model support and local deployment.
selfgpt — Tests how well different language models can simulate human user messages.
KernelAI — Runs 53 open-weight models offline with no account or subscription required.
HuggingChat UI — It provides a customizable open source chat interface for any OpenAI-compatible backend.
RWKV — Combines linear-time RNN efficiency with transformer-level performance.
Sidekick — Chat with local models using your own files and websites offline.
Uniichat — Maintains one unending chat with a memory tree that never loses historical detail.

All MOSS alternatives