OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Moss

Moss is an open-source project providing MOSS-003 conversational language models, dialogue datasets, and code for local deployment and fine-tuning. Its models support Chinese and English dialogue; a plugin-tuned variant is described for search, text-to-image generation, calculation, and equation solving.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Moss — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Moss when you want model files and development resources for local conversational-AI prototyping or evaluation, and can provide the required GPU memory and deployment setup.

Who it is for

  • Developers prototyping Chinese- or English-language dialogue applications.
  • Researchers evaluating conversational text generation or plugin-assisted dialogue.

Documented uses

  • Build a conversational assistant using a MOSS model.
  • Prototype multi-turn dialogue that can use documented plugin types.
  • Evaluate model responses on natural-language generation tasks.
  • Fine-tune a model using the project’s documented code and dialogue data.
Before you choose

Requirements & limits

The repository is public and identifies the project as open source. Code, data, and model assets have separate stated licenses: Apache 2.0, CC BY-NC 4.0, and GNU AGPL 3.0, respectively.

  • The MOSS-Moon model family is described as having 16 billion parameters.
  • The repository estimates batch-size-one GPU memory use at 31 GB to load FP16, 16 GB to load INT8, and 7.8 GB to load INT4; estimates rise with conversation length.
  • For a conversation round, the memory estimates are 42 GB for FP16, 24 GB for INT8, and 12 GB for INT4; at the maximum 2,048-token conversation length they are 81 GB, 46 GB, and 26 GB, respectively.
  • Quantized models are documented as not supporting model parallelism.
  • The maintainers caution that generated output can contain factual errors, bias, or harmful content.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Prepare a local MOSS environment

Clone the MOSS repository and create its documented Python 3.8 Conda environment with dependencies installed.

Before you start

  • Conda
  • Git
  • Access to a local machine or remote server

Steps

  1. Clone the repository with the command below.

    Documented command
    git clone https://github.com/OpenMOSS/MOSS.git

  2. Change into the repository directory with cd MOSS.

  3. Create and activate the environment with conda create --name moss python=3.8 and conda activate moss.

  4. Install the listed dependencies with the command below.

    Documented command
    pip install -r requirements.txt

Expected outcome

The repository is cloned and the `moss` Conda environment is prepared with the requirements installed.

Guide limitations

  • These steps cover setup only; they do not document model download or inference execution.
  • The README lists inference memory requirements by quantization and conversation length; verify available GPU memory before proceeding to the local deployment usage instructions.
  • Quantized models do not support model parallelism.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
14 October 2026
Original website title
OpenMOSS/MOSS · Official homepage title · retained in its source language
How we review sourcesSuggest a correction