Skip to content
@zenlm

Zen LM

Frontier language models from 600M to 480B parameters — edge to cloud. Open weights, Zen architecture.
README.md

Zen LM

Frontier language models from 600M to 480B parameters. Open-weight models optimized for edge devices through cloud-scale deployments, built on the Zen architecture with efficient inference via Rust, MLX, and GGUF.

Zen LM develops state-of-the-art language models spanning ten modalities from 600M (embedded) to 480B (frontier research). All models use the Zen architecture with RoPE embeddings, SwiGLU activation, grouped-query attention, and Flash Attention 2. Available through the Hanzo API, Hugging Face, vLLM, and local inference via MLX/GGUF.

Quick Start

# Using the Hanzo API
from hanzo import Client

client = Client()
response = client.chat.completions.create(
    model="zenlm/zen4-pro",
    messages=[{"role": "user", "content": "Hello!"}]
)

# Using Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained("zenlm/zen4-pro")

Model Family

Language Models

Model Parameters Context Use Case
zen-coder-480b 480B (MoE) 256K Frontier code
zen4-max 397B (MoE) 128K Frontier reasoning
zen-max 235B (MoE) 128K Production flagship
zen4-thinking 80B (MoE) 128K Deep reasoning
zen4-coder-pro 80B (MoE) 128K Professional code
zen-voyager 32B 128K Balanced
zen4-mini 8B 128K Fast inference
zen-eco 4B 32K Edge devices
zen-nano 0.6B 8K Embedded

Multimodal

Model Type Description
zen-omni Audio+Vision Hypermodal 30B MoE
zen-vl Vision Vision-language models (4B/8B/30B)
zen-video Video High-quality video synthesis
zen-scribe Speech Speech recognition
zen-translator Translation Real-time multimodal translation

Core Projects

Project Description
zen Zen AI model family -- efficient models for edge and cloud
engine High-performance inference engine -- Rust/MLX/GGUF
gym Unified fine-tuning for 100+ LLMs and VLMs
zen-omni Zen-Omni 30B -- hypermodal AI with MLX/GGUF
docs Documentation and model cards

Ecosystem

Organization Focus Link
Hanzo AI The Open AI Cloud: Enso, agents, MCP, inference github.com/hanzoai
Lux Network Post-quantum blockchain, FHE, multi-consensus github.com/luxfi
Zen LM Frontier language models, 600M-480B parameters github.com/zenlm
Zoo Labs Open AI research, DeSci, 501(c)(3) foundation github.com/zoo-labs
Lux FHE Fully homomorphic encryption research + SDK github.com/luxfhe
Lux C++ High-performance C++ libraries github.com/luxcpp
Hanzo KMS Enterprise secret management github.com/hanzokms
Hanzo ID Identity, OAuth2/OIDC, WebAuthn github.com/hanzoid

Links

  • Papers -- 329+ research papers
  • Stats -- 38,906+ commits, 59M net LOC
  • Security -- cryptographic audit trail
  • History -- 2008-2026 timeline

Resources

Apache 2.0 for code, model-specific licenses for weights.

Co-developed by Hanzo AI and Zoo Labs

Popular repositories Loading

  1. engine engine Public

    High-performance inference engine — Rust/MLX/GGUF

    Rust 6 1

  2. zen-dub zen-dub Public

    Zen Dub - Real-time high-quality lip-sync model

    Python 3 1

  3. zen5-engine zen5-engine Public

    Zen5 inference engine. Fork of antirez/ds4 (DeepSeek V4 Flash native runtime) rebranded for the Zen LLM family.

    C 3

  4. vjepa2 vjepa2 Public

    Forked from facebookresearch/vjepa2

    PyTorch code and models for VJEPA2 self-supervised learning from video.

    Python 2

  5. zen zen Public

    Zen AI model family — efficient models for edge and cloud

    MDX 2

  6. enso enso Public

    Enso Diffusion: Multimodal Mixture of Unbound Experts (MUEN) — unifying text, vision, and audio through diffusion-based intelligence.

    Python 1

Repositories

Showing 10 of 90 repositories
View all repositories

Top languages

Loading…

Most used topics

Loading…