Zen MoDE — Mixture of Distilled Experts

Zen LM

Frontier AI models for code, reasoning, vision, video, audio, 3D, and agentic workflows

49 models across 10 modalities. 15 production API models from 4B to 1T+ parameters. Open weights on HuggingFace. OpenAI-compatible API. Built by Hanzo AI (Techstars '17).

Try Zen Free Documentation API Reference HuggingFace

Models

1T+

Max Parameters

262K

Max Context

Modalities

$0.30

From $/MTok

Flagship Models

Three tiers — from efficient edge to trillion-parameter frontier scale

FRONTIER

zen4-max

1.04T MoE · 32B active · 256K ctx

Trillion-parameter frontier model with deep reasoning. Our largest and most capable model.

FLAGSHIP

zen4

744B MoE · 40B active · 202K ctx

Flagship MoE for complex reasoning and multi-domain tasks. Optimal intelligence-to-speed ratio.

CODE

zen4-coder

480B MoE · 35B active · 262K ctx

Code-specialized MoE model for generation, review, debugging, and agentic programming.

View all models

API Pricing

Pay-as-you-go. $5 free credit on signup. No minimum commitment.

FREE TIER

Free credit on every new account

All 15 API models
OpenAI-compatible API
30-day expiry

Get Started Free

PAY AS YOU GO

$0.30/MTok

Starting from — scale with usage

Prepaid credits
Real-time usage tracking
No surprise bills

Add Credits

ENTERPRISE

Custom

Volume pricing, SLAs, dedicated support

Volume discounts
Dedicated infrastructure
SLA guarantees

Contact Sales

Zen4 Generation — Production API

Model	Architecture	Context	Tier	Input $/MTok	Output $/MTok
zen4-max	1.04T (32B active) MoE	256K	ultra max	$3.60	$3.60
zen4	744B (40B active) MoE	202K	ultra max	$3.00	$9.60
zen4-ultra	744B (40B active) MoE + CoT	202K	ultra max	$3.00	$9.60
zen4-pro	80B (3B active) MoE	131K	ultra	$2.70	$2.70
zen4-thinking	80B (3B active) MoE + CoT	131K	pro max	$2.70	$2.70
zen4-coder	480B (35B active) MoE	262K	ultra	$3.60	$3.60
zen4-coder-pro	480B Dense BF16	262K	ultra max	$4.50	$4.50
zen4-coder-flash	30B (3B active) MoE	262K	pro max	$1.50	$1.50
zen4-mini	8B Dense	40K	pro	$0.60	$0.60

Zen3 Generation — Specialized Models

Model	Architecture	Context	Tier	Input $/MTok	Output $/MTok
zen3-omni	~200B Dense Multimodal	202K	pro max	$1.80	$6.60
zen3-vl	30B (3B active) MoE VL	131K	pro max	$0.45	$1.80
zen3-nano	4B Dense	40K	pro	$0.30	$0.30
zen3-guard	4B Dense	40K	pro	$0.30	$0.30
zen3-embedding	Embedding (3072 dim)	8K	pro max	$0.39	—

100 chat messages

zen4-mini · 50K in + 50K out

$0.06

1K code completions

zen4-coder-flash · 500K in + 500K out

$1.50

10K embeddings

zen3-embedding · 1M input

$0.39

Heavy daily use

zen4 · 1M in + 1M out

$12.60

Full pricing details

Quick Start

Install the Hanzo SDK — supports OpenAI and Claude-style endpoints, plus 100+ providers

Python — Hanzo SDK

pip install hanzoai

from hanzoai import Hanzo

client = Hanzo(api_key="hk-your-api-key")

response = client.chat.completions.create(
    model="zen4",
    messages=[
        {"role": "user", "content": "Hello, Zen."}
    ],
)
print(response.choices[0].message.content)

TypeScript — Hanzo SDK

npm install hanzoai

import Hanzo from "hanzoai";

const client = new Hanzo({
  apiKey: "hk-your-api-key",
});

const response = await client.chat.completions.create({
  model: "zen4-coder",
  messages: [
    { role: "user", content: "Write a React hook" }
  ],
});
console.log(response.choices[0].message.content);

curl

curl https://api.hanzo.ai/v1/chat/completions \
  -H "Authorization: Bearer hk-your-api-key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "zen4",
    "messages": [
      {"role": "user", "content": "Hello, Zen."}
    ]
  }'

Streaming (Python)

from hanzoai import Hanzo

client = Hanzo(api_key="hk-your-api-key")

stream = client.chat.completions.create(
    model="zen4-max",
    messages=[
        {"role": "user", "content": "Explain MoE"}
    ],
    stream=True,
)
for chunk in stream:
    if chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="")

SDK Features

Multi-SDK — Python, TypeScript, Go, Rust

OpenAI Compatible — drop-in replacement

100+ Providers — Zen + Claude + GPT + more

Streaming — SSE, async, batch

Python SDK

pip install hanzoai

TypeScript SDK

npm install hanzoai

Go SDK

go get github.com/hanzoai/go-sdk

Rust SDK

cargo add hanzoai

Full Model Library

49 models across 10 categories — text, code, vision, video, audio, 3D, safety, embeddings, and agents

Foundation

Core models from 0.6B edge to 1T+ frontier scale

6 models

zen-nano

0.6B Dense · 32K ctx

Ultra-lightweight edge model — 44K tokens/sec, fits in 0.4GB

HuggingFace

zen-eco

4B Dense · 32K ctx

Efficient general-purpose — 33K tokens/sec, 2–8GB memory

HuggingFace

zen

8–32B Dense · 32K ctx

Standard foundation — versatile for most tasks

HuggingFace

zen-pro

32B Dense · 32K ctx

Professional grade — 19K tokens/sec, high-quality outputs

HuggingFace

zen-max

1.04T MoE · 256K ctx

Maximum scale — same model as zen4-max, open weights

HuggingFace

zen-nextPREVIEW

TBD Dense · 256K ctx

Next-generation preview

Zen4 Generation

Latest generation — 10 production API models

10 models

zen4-max

1.04T (32B active) MoE · 256K ctx

Frontier scale — deepest reasoning capabilities

HuggingFace

zen4

744B (40B active) MoE · 202K ctx

Flagship intelligence — multi-domain reasoning

HuggingFace

zen4-ultra

744B (40B active) MoE + CoT · 202K ctx

Maximum reasoning with chain-of-thought

HuggingFace

zen4-pro

80B (3B active) MoE · 131K ctx

High capability — efficient MoE architecture

HuggingFace

zen4-thinking

80B (3B active) MoE + CoT · 131K ctx

Deep reasoning — chain-of-thought enabled

zen4-mini

8B Dense · 40K ctx

Ultra-fast inference — cost-effective

HuggingFace

zen4-coder

480B (35B active) MoE · 262K ctx

Code generation — 262K context for large codebases

HuggingFace

zen4-coder-pro

480B Dense BF16 · 262K ctx

Premium code — full-precision dense model

HuggingFace

zen4-coder-flash

30B (3B active) MoE · 262K ctx

Fast code — low-latency completions

HuggingFace

zen4-coder-nextPREVIEW

TBD MoE · 262K ctx

Next-gen code generation — preview

Code

Specialized for generation, review, debugging, and agentic programming

3 models

zen-coder

32B Dense · 131K ctx

Multi-language code generation and understanding

HuggingFace

zen-coder-flash

7B Dense · 32K ctx

Low-latency code completions

HuggingFace

zen-code

14B Dense · 32K ctx

Legacy code model — still available on HuggingFace

HuggingFace

Vision & Image

Image understanding, generation, editing, and design

5 models

zen-vl

32B Dense Multimodal · 32K ctx

Vision-language understanding — image analysis

HuggingFace

zen-omni

72B Dense Multimodal · 131K ctx

Hypermodal — text, vision, audio, code in one model

HuggingFace

zen-artist

Image Generation

High-resolution image generation, multiple styles

zen-artist-edit

Image Editing

Edit-by-instruction, inpainting, outpainting

zen-designerSOON

Design Generation

UI/UX design, graphics, mockups

Video

Video generation, understanding, and world modeling

4 models

zen-directorSOON

Text-to-Video

Cinematic-quality text-to-video generation

zen-videoSOON

Video Understanding

Video analysis, frame-by-frame understanding

zen-video-i2vSOON

Image-to-Video

Animate images into video sequences

zen-voyagerSOON

World Model

Spatial reasoning and world simulation

Audio & Speech

Music generation, voice synthesis, transcription, and translation

7 models

zen-scribe

Speech-to-Text

Multi-language transcription

zen-translator

Translation

100+ languages, context-aware translation

zen-dubSOON

Voice Synthesis

Multi-language voice dubbing and synthesis

zen-dub-liveSOON

Real-time Voice

Ultra-low latency real-time voice

zen-musicianSOON

Music Generation

Multi-instrument music composition

zen-foleySOON

Sound Effects

Text-to-SFX, foley art generation

zen-liveSOON

Real-time Translation

Bidirectional real-time speech translation

3D & Spatial

3D asset generation and world simulation

2 models

zen-3dSOON

3D Generation

Text-to-3D and image-to-3D asset generation

zen-worldSOON

World Simulation

Spatial reasoning and environment simulation

Safety & Guardrails

Content moderation and safety classification

3 models

zen-guard

8B Dense · 32K ctx

Content moderation and safety classification

HuggingFace

zen-guard-gen

8B Dense · 32K ctx

Safe generation with built-in guardrails

zen-guard-stream

4B Dense · 8K ctx

Low-latency streaming moderation

Embedding & Retrieval

Text embeddings and search reranking

2 models

zen-embedding

3072 dim · 8K ctx

High-dimensional text embeddings

HuggingFace

zen-reranker

568M Dense · 8K ctx

Cross-encoder search reranking

HuggingFace

Agents

Agentic AI with tool use and multi-step planning

1 models

zen-agentPREVIEW

32B Dense · 131K ctx

Agentic AI — tool use, multi-step planning, autonomous execution

10 Modalities

One model family covering every AI capability

Text

14 models

Chat, reasoning, analysis

Code

9 models

Generation, review, debugging

Vision

5 models

Understanding, generation, editing

Video

4 models

Generation, understanding, I2V

Audio

7 models

Speech, music, translation

3D

2 models

Generation, world simulation

Safety

3 models

Moderation, guardrails

Embedding

2 models

Search, retrieval, RAG

Agents

1 models

Tool use, planning

Math

6 models

Reasoning, proof, computation

Architecture

Zen MoDE — curating best open-source foundations and fusing them into a unified, high-performance family

Consumer Line

Dense and MoE models from 4B to 80B. Edge-deployable dense models and efficient MoE flagships with only 3B active parameters.

Coder Line

Code-specialized models trained on 8.47B tokens of real agentic programming data. Fast completions to full-precision code intelligence.

Ultra Line

Trillion-parameter MoE models for cloud deployment. 1.04T parameters with 32B active for frontier-scale reasoning.

Efficient MoE

Mixture-of-Experts delivers frontier performance with only 3B active parameters — runs on consumer hardware.

Long Context

Up to 262K context on code models, 256K on frontier models. Dense models support 32–40K for efficient local inference.

Zen MoDE

Mixture of Distilled Experts — curating the best open-source foundations and fusing into a unified model family.

Zen Agentic Dataset

8.47 billion tokens of real-world agentic programming — not synthetic data

8.47B

Training Tokens

3.35M

Training Samples

1,452

Repositories

15yr

History (2010–2025)

48%

Git History

4.03B tokens

29%

Agentic Debug Sessions

2.42B tokens

13%

Architecture Discussions

1.14B tokens

10%

Code Review Sessions

0.86B tokens

Dataset documentation

Open Weights

Download, self-host, and fine-tune — multiple formats for every platform

SafeTensors

Full precision for HuggingFace Transformers

GGUF

Quantized for llama.cpp / Ollama

MLX

Apple Silicon optimized

ONNX

Cross-platform inference

Browse All on HuggingFace

Hanzo Ecosystem

Zen models power the entire Hanzo AI platform

Research

130+ technical papers across AI, cryptography, and consensus

Build with Zen LM

49 models. 10 modalities. Open weights. From $0.30/MTok.

Try Zen Free Documentation Get API Key

Try Zen

Zen LM

Flagship Models

zen4-max

zen4

zen4-coder

API Pricing

Zen4 Generation — Production API

Zen3 Generation — Specialized Models

Quick Start

SDK Features

Python SDK

TypeScript SDK

Go SDK

Rust SDK

Full Model Library

Foundation

Zen4 Generation

Code

Vision & Image

Video

Audio & Speech

3D & Spatial

Safety & Guardrails

Embedding & Retrieval

Agents

10 Modalities

Text

Code

Vision

Video

Audio

3D

Safety

Embedding

Agents

Math

Architecture

Consumer Line

Coder Line

Ultra Line

Efficient MoE

Long Context

Zen MoDE

Zen Agentic Dataset

Git History

Agentic Debug Sessions

Architecture Discussions

Code Review Sessions

Open Weights

SafeTensors

GGUF

MLX

ONNX

Hanzo Ecosystem

Hanzo Chat

Hanzo Cloud

Hanzo MCP

Hanzo Dev

LLM Gateway

Hanzo Industries

Research

Zen Technical Report

Zen Coder

Zen Agent

Zen Guard

Zen Omni

Zen Artist

Zen Pro

Zen MoDE

Build with Zen LM