Frontier AI families
Proprietary and open weight models for text, decisions, video, image, and more — one account, one API key.
Live from api.hanzo.aiHanzo's proprietary frontier model — Opus-class reasoning by default with 1M-context overflow.
Three families
Enso routes. Kai decides. Zen reasons.
- routesEnsoPicks the model for every call, in every modality.Explore Enso →
- decidesKaiAnswers a choice, a score or a yes/no, with the probability of each.11 · Suites won · Kai vs JevExplore Kai →
- reasonsZenOpen weights for the code, the copy, the images and the voice.Zen 6 and Zen 6 Flash now · Zen 7 in research previewExplore Zen →
Enso vs Zen
Proprietary orchestration, on open foundations
Enso
Proprietary · Hanzo Cloud only- Flash, Pro and Ultra presets
- OpenAI- and Anthropic-compatible
- Metered and audited per organization
Zen
Open weights · Zoo Labs Foundation- Free to download and self-host
- Or call it on Hanzo Cloud
- From a 501(c)(3) non-profit
Zen models
Open weights, made with Zoo Labs Foundation. Download every Zen model at huggingface.co/zenlm, or call it here.
Canonical Zen5 default. a 35B frontier MoE (3B active) (35B total / 3B active per token, released Apr 2026, Apache-2.0). 256K context, agentic-trained, OpenAI + Anthropic API. The everyday Zen5 chat model.
Zen Flash IQ2_XXS-imatrix weights (81 GB GGUF on zenlm/zen-5-pro-gguf). 284B total / 37B active per token, 1M context, asymmetric routed-MoE quant. Fits a single 128 GB Apple Silicon / DGX Spark / H100 80 GB.
Code-specialized Zen5 tier. 80B sparse MoE tuned for repo-scale code understanding, agentic refactoring, and tool-use coding loops.
Smallest and cheapest text-only Zen5 chat tier. 4B-class dense, sub-100ms TTFT, 32K context. For high-volume routing and simple agent loops.
Reads images and reasons over them — visual Q&A, document and chart understanding, grounded captioning.
Dense text embeddings for semantic search, retrieval, and clustering.
The next open-weight generation after Zen 6, in research preview.
Kai decision models
Hanzo Kai is a structured decision model answering typed questions at /v1/decisions with calibrated probabilities and zero output tokens. Full multimodal decision architecture (~850M parameters, 32K reach); native vision and audio towers: Kai-1.7 target.
Ahead of Jev on 11 suites: AG News, DAIR Emotion, Banking77, Support triage, Email spam, Phishing, Toxicity, RAG relevance, Model routing, Typed decisions, MASSIVE, 51 languages. 84.4% mean accuracy across these 11 suites, against Jev's 76.6%. 100.0% of repeated requests answer identically, against Jev's 44.2%.
Creative models
Video, images, 3D, sound effects, voice and music, each its own model, in research preview.
Cinematic video generation, character continuity, motion, animation, and shot-level creative control.
High-fidelity image generation, creative direction, editing, character design and consistent visual assets.
3D assets, geometry, materials, environments and spatially consistent scenes.
Film Foley, immersive sound effects, environmental audio and synchronized scene sound.
Directable speech synthesis, character performances, dubbing and multilingual dialogue.
Soundtracks, original songs, musical ideas, cue-driven scoring and creative music production.
Flagships from other labs
Claude Fable 5 is a Mythos-class model from Anthropic, built for autonomous knowledge work and coding. It supports text, image, and file inputs with text output, with reasoning support and...
GPT-3.5 Turbo is OpenAI's fastest model. It can understand and generate natural language or code, and is optimized for chat and traditional completion tasks. Training data up to Sep 2021.
Gemini 2.5 Pro is Google’s state-of-the-art AI model designed for advanced reasoning, coding, mathematics, and scientific tasks. It employs “thinking” capabilities, enabling it to reason through responses with enhanced accuracy...
DeepSeek-V3 is the latest model from the DeepSeek team, building upon the instruction following and coding abilities of the previous versions. Pre-trained on nearly 15 trillion tokens, the reported evaluations...
Grok 4.20 is a reasoning model from SpaceXAI with industry-leading speed and agentic tool calling capabilities. It combines the lowest hallucination rate on the market with strict prompt adherance, delivering...
Muse Glimmer 30B is a dense, open-weight multimodal model from Meta Superintelligence Labs, distilled from Muse Spark and optimized for autonomous agents on consumer hardware. It is suited for long-horizon...
Every model, and what it costs
Dollars per million tokens, in and out, from the live catalog.
Try any model
Test any model directly on the gateway. Switch models by changing one string.
Sign in to chat with Enso, Zen, or any model on the endpoint. Pick a model, start a thread, billed to your account and kept private.
Your first request
The chat-completions shape you already send, at api.hanzo.ai/v1, in the language you write.
curl https://api.hanzo.ai/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer $HANZO_API_KEY" \
-d '{
"model": "enso",
"messages": [{"role": "user", "content": "Hello!"}]
}'from openai import OpenAI
client = OpenAI(
api_key=os.environ["HANZO_API_KEY"],
base_url="https://api.hanzo.ai/v1"
)
response = client.chat.completions.create(
model="enso",
messages=[{"role": "user", "content": "Hello!"}]
)
print(response.choices[0].message.content)import OpenAI from 'openai'
const client = new OpenAI({
apiKey: process.env.HANZO_API_KEY,
baseURL: 'https://api.hanzo.ai/v1'
})
const response = await client.chat.completions.create({
model: 'enso',
messages: [{ role: 'user', content: 'Hello!' }]
})
console.log(response.choices[0].message.content)package main
import (
"context"
"fmt"
"os"
"github.com/sashabaranov/go-openai"
)
func main() {
cfg := openai.DefaultConfig(os.Getenv("HANZO_API_KEY"))
cfg.BaseURL = "https://api.hanzo.ai/v1"
client := openai.NewClientWithConfig(cfg)
resp, _ := client.CreateChatCompletion(context.Background(),
openai.ChatCompletionRequest{
Model: "enso",
Messages: []openai.ChatCompletionMessage{
{Role: openai.ChatMessageRoleUser, Content: "Hello!"},
},
},
)
fmt.Println(resp.Choices[0].Message.Content)
}Related capabilities