18+ Zen Models

Hanzo Zen Models
Language · Code · Vision · Audio

18+ open-weight models from 0.6B to 1T+. Zen MoDE (Mixture of Distilled Experts) architecture. All models on HuggingFace.

zen-models
18+
Models
1.04T
Max Params
2M
Max Context
From
Language + Code + Vision + Audio
Zen MoDE — Mixture of Distilled Experts
OpenAI-compatible API
Open-weight on HuggingFace

Why Zen Models?

OpenAI-Compatible API

Drop-in replacement for GPT/Claude. Same SDK, same format, lower cost.

Open Weight

All Zen models are open-weight on HuggingFace. Run them yourself.

Full Spectrum

0.6B edge to 1T+ frontier. Text, code, vision, audio — one API.

MoE Efficiency

Mixture-of-Experts activates only a fraction of params per token.

Every Modality

Text, vision, code, speech — all through a single unified API

Language

Chat, reasoning, RAG

zen4-max
zen4-ultra
zen4-pro

Code

Generation, review, debug

zen4-coder-flash
zen4-coder
zen4-mini

Vision

Image, video, OCR

zen-vl-4b
zen-vl-8b
zen-vl-30b

Audio

Speech-to-speech <300ms

zen-omni
Coming Soon
Next Generation

Zen 5 Ultra

2T+ parameter MoDE. Trained on-chain via NVIDIA TEE confidential compute on hanzo.network. The largest open-weight model in history.

2T+
Parameters
MoDE
Mixture of Distilled Experts
TEE
On-Chain Verifiable Training

Ready to build with Zen?

18+ models, full spectrum of modalities, one API. Start building today.