Moxin LM
Open Source Foundation Models

From reasoning to speech, our models are designed for the next generation of human-computer interaction.

Visit Moxin LM Hugging Face Page

Learn more about our foundation models and research.

huggingface.co/moxin-org
Open Creation Efficient Deployment Voice

Open Creation

The Moxin-7B series is our truly open, SOTA-performing LLM and VLM. We build, fine-tune, and openly release our own models, ensuring complete reproducibility and transparency.

Moxin-7B-LLM

Our flagship general-purpose model. Fine-tuned for instruction following, coding, and reasoning.

7B
Params
32k
Context
SOTA
Perf
HF Learn More GitHub Paper

Moxin-7B-VLM

Vision-Language Model capable of understanding images, charts, and diagrams with high precision.

HF Learn More GitHub Paper

Efficient Deployment

We specialize in extreme quantization, creating resource-efficient variants of popular models (like DeepSeek and Kimi) to run anywhere. We unleash the power of reproducible AI πŸš€.

Kimi K2 Thinking

Optimized GGUF version of Kimi K2 Thinking model.

HF Learn More

MiniMax M2

Efficient GGUF quantization for MiniMax M2.

HF Learn More

Qwen3 Next 80B

A3B Instruct GGUF version of Qwen3 Next 80B.

HF Learn More

Qwen3 235B

Massive 235B parameter model quantized for deployment.

HF Learn More

DeepSeek V3

Latest DeepSeek V3 model optimized for Moxin.

HF Learn More

GLM 4.6

General Language Model 4.6 GGUF quantization.

HF Learn More

DeepSeek R1

Reasoning model optimized for efficient deployment.

HF Learn More

Voice

Moxin Voice brings speech synthesis, voice cloning, and speech recognition into a single fully local workflow for desktop products, edge devices, and research environments.

NEW
Runs locally 5-10s cloning 3-10 min training 14+ preset voices

Moxin Voice

Text-to-speech and automatic speech recognition that run entirely on-device, with no cloud API dependency and a strong focus on speed, privacy, and flexible deployment.

Human-like text-to-speech and automatic speech recognition running entirely on-device

Recording, preview, and WAV export built into a workflow that is easy to demo and integrate

14+ preset voices with room to extend, fine-tune, and train custom voice profiles

Well-suited for desktop apps, edge devices, voice-enabled agents, and research workflows

View on GitHub

Voice Sample

WAV

Moxin Voice Vivian sample

Zero-Shot Voice Cloning

Create a usable voice profile from just 5-10 seconds of reference audio for fast demos and personalized speech experiences.

Few-Shot High-Fidelity Training

Train a more stable and accurate custom voice model with 3-10 minutes of audio when quality matters most.

Local TTS + ASR

Keep text-to-speech, automatic speech recognition, recording, playback, and export inside a fully local pipeline.

Native Performance Stack

Built with Rust, Makepad, and GPT-SoVITS v2 to optimize privacy, responsiveness, and deployment flexibility.

Build with Moxin LM

Robotics & Automation

Fine-tune for specific robotics commands and industrial applications.

Edge AI Solutions

Run AI directly on devices for privacy-first, low-latency applications.

Research Platform

Ideal for academic research with full reproducibility and transparency.