Join the Hanzo beta and get 50% off any paid plan

Smarter, Faster, Open Source. Pick all three.

Frontier intelligence up to 200× faster and 400× cheaper. Every model behind one key.

By continuing, you agree to Hanzo's Terms and Privacy Policy.

hanzo.ai/chat

Hanzo AI

Recents

Launch plan

Weekly report

▍

9:41

Show me the build

• Wrote app/agents.tsx• Wrote lib/keys.ts• Ran pnpm test └ pass ⚡

Ask anything

maxpower

Claude Code

Zen6116/s

Codex

Zen6118/s

Gemini CLI

Opus 4.874/s

Hanzo Dev

Opus 4.8107/s

$0.0000demo

9:41

▍

maxpower

Claude Code

Zen6116/s

Codex

Zen6118/s

Gemini CLI

Opus 4.874/s

Hanzo Dev

Opus 4.8107/s

$0.0000demo

9:41

Show me the build

• Wrote app/agents.tsx• Wrote lib/keys.ts• Ran pnpm test └ pass ⚡

Ask anything

maxpower

Claude Code

Zen6116/s

Codex

Zen6118/s

Gemini CLI

Opus 4.874/s

Hanzo Dev

Opus 4.8107/s

$0.0000demo

Backed by

Hanzo.Apps

One account. Every tool you build with.

How do black holes form?

A massive star runs out of fuel and gravity collapses its core into a singularity.

What causes the aurora?

Solar particles hit gases high over the poles and make them glow.

Chathanzo.ai/chat Explore

dev — projects/main

$ dev “Require a token on every API route”● Read src/middleware/auth.ts● Edit src/middleware/auth.ts function guard(req) {- return next(req)+ const token = bearer(req)+ if (!token) return deny()+ req.user = await verify(token)+ return next(req)● Run pnpm test✓ Tests pass

Devhanzo.ai/dev Explore

Routine · Month-end close

Read the month’s card charges

Matched each charge to its receipt

Asked about the ones with no receipt

Filed the month-end report

Done. 2 questions waiting for you.

Bothanzo.bot Explore

IroImage

KomaVideo

Mori3D

OtoSound

KoeVoice

UtaMusic

Studiohanzo.ai/studio Explore

# launch

Kaiagent

Tests pass on main. Deploy to production?

Grace

Approved. Ship it.

Sam

Release notes are in the doc.

Ada

Live. Thanks, everyone.

Teamhanzo.team Explore

Hanzo.Models

Frontier AI for everything you build.

Language, routing, decisions and creative models.

Flagship model

Enso

Orchestrates every model for higher intelligence, faster and at lower cost.

RoutingAgentsMulti-modal

Kai

Precise decisions and structured reasoning.

Logic

Zen

Deep reasoning for complex problems.

Reasoning

Iro

Image generation and visual understanding.

Images

Koma

Video generation and world simulation.

Video

Oto

Sound generation and audio understanding.

Audio

Koe

Voice synthesis and speech understanding.

Voice

Uta

Music generation and composition.

Music

Mori

3D generation and spatial understanding.

Hanzo.Smarter

Pareto Frontier of Intelligence

GPQA Diamond accuracy against the cost of 1,000 questions.

How Enso routes Route your own requests with Enso

Hanzo.Faster

Faster where the work happens.

Hanzo Dev and Kai, each measured side by side.

Hanzo Dev vs Claude Code

3.02×faster on the same coding tasksPaired over 42 tasks both agents solved · 95% interval 2.06–4.36×.

Median time a task · less time is faster

Claude Code9.52 s

Hanzo Dev2.29 s

Every coding task and both runs

Kai vs an LLM call

9.6×faster than an LLM call, over the networkEach a p50 timed at the client, network included: a Claude Sonnet 5.5 call as Claude Code saw it, over 650 calls, and a Kai decision on api.hanzo.ai, over 100. Served on one CPU with no network, its input repeated, Kai answers in 24 ms.

One call · less time is faster

Claude Sonnet 5.5 call2,307 ms

Kai on api.hanzo.ai241 ms

Kai served on one CPU, no network24 ms

How Kai makes decisions

Install Hanzo Dev Ask Kai your first decision

Hanzo.Cheaper

The same intelligence. A smaller bill.

Enso routes each answer to the right model. Kai answers a decision directly.

Enso · measured on the same questions

1.6× lessEnso Flash against the best single model, GPQA DiamondEnso Flash at 90.6% and $3.26 per 1,000 new questions, against 93.1% at $5.16 for the best single model. Both research runs.

How Enso lowers the cost of an answer

Kai vs frontier LLMs

up to 530× lessthan frontier LLMs, per input tokenBilled per million tokens, from the model list. Kai bills $0.021 for input and nothing for output; each LLM here also bills every token of the answer it writes.

Per million tokens, input and output · less is cheaper

Claude Fable 5.1$11.13 in · $55.65 out

GPT-6 Astra$11.13 in · $55.65 out

GPT-5.5$5.57 in · $33.39 out

Claude Opus 5$5.57 in · $27.83 out

Kai$0.021 in · $0 out

How Kai makes decisions

Hanzo.Bill

The model you pay for today

Tokens a month

What they are

100M tokens a month at each list rate · less is cheaper

Claude Opus 5$1,113

Enso Flash$66

Enso$66

Enso Pro$880

Enso Ultra$1,100

Claude Opus 5 bills $1,113 a month for these tokens. Every rate is the model list’s, per million tokens in and out; which tier answers as well is the chart above.

Which Enso tier for which job Every model rate

Hanzo.Source

We ship every day.

The code and the commit record are public.

Commits, past year

91,401

Commits, past 7 days

345

Original public repositories

1,617

69 organizations checked · non-fork repositories

Active this month

repositories with a commit in 30 days

GitHub stars

122

as GitHub counts them

GitHub forks

133

as GitHub counts them

Commits a day

Commits a day, from Oct 31, 2025 to Oct 30, 2026

Latest commits

hanzoai/guishell: the wordmark's files in oxfmt's formOct 30, 06:04 UTC
hanzoai/appChrome: the name lights with the bar's highlight, never an outline (0.2.56)Oct 30, 04:42 UTC
hanzoai/cisite: the release's record of itself goes after its pagesOct 30, 04:23 UTC
hanzoai/uifooter: the legal line beside the copyright, the wordmark at the row's end, the panel's ground; 8.29Oct 30, 03:57 UTC
hanzoai/aimodels: ?id= narrows the listing to the models a page namesOct 30, 03:30 UTC
hanzo-docs/docssdks: @hanzo/app and @hanzo/mod, what each shipsOct 30, 03:28 UTC
hanzoai/benchmarksrecord: add Kai-1.9 Cand29 joint coherence capability benchmark resultsOct 30, 01:08 UTC
hanzoai/botchore(license): finish Apache-2.0 documentation and desktop attributionOct 30, 00:53 UTC

Public GitHub snapshot · read Oct 30, 06:13

Read the source

What it is written in

Python 22.2%

TypeScript 21.4%

Go 18.1%

Shell 12.2%

JavaScript 9.3%

Other 16.9%

The primary language of each of 1,617 original repositories, as GitHub reads it

hanzoai/aiHanzo AI — LLM control plane, RAG, and model hub. Native Go routing for 66+ models over an OpenAI-compatible /v1 API; mounts as the 'ai' subsystem inside hanzoai/cloud.

51 contributors

1,607 commits this yearPushed Oct 30, 2026

api.hanzo.ai

hanzoai/engineHanzo AI native inference engine - Rust-based LLM & embedding engine for foundational models

Rust

82 contributors

1,618 commits this yearPushed Oct 29, 2026

hanzoai/cloudHanzo Cloud — OSS core. One Go binary. Dual-licensed Apache-2.0 OR MIT.

1 contributor

123 commits this yearPushed Sep 30, 2026

hanzoai/devHanzo Dev — fast, local AI coding agent for your terminal (Rust CLI, powered by Hanzo AI; Apache-2.0).

Python

154 contributors

206 commits this yearPushed Oct 29, 2026

Forge

hanzoai/cliThe Hanzo CLI — an AI coding agent and every product of the Hanzo cloud, in one Rust binary. curl -fsSL https://hanzo.sh | sh

Rust

2 contributors

401 commits this yearPushed Oct 29, 2026

hanzo.sh

hanzoai/mcpModel Context Protocol server with 260+ tools for AI agents

Rust

2 contributors

192 commits this yearPushed Oct 29, 2026

Forge hanzo.ai

Hanzo Forge

Our own Git forge, git.hanzo.ai. Its figures are its own, read from it signed out, and never added to the GitHub figures.

461

public repositories, open to read signed out

145

of them mirror our repositories on GitHub, Hanzo, Lux and Zoo together

243

community projects published there

Forge listing · read Oct 30, 2026

Explore the forge

Explore every repository

Hanzo.Agents

Works with your favorite agents.

Claude Code, Codex, Cursor and the rest, every model behind one key. Or let your agent sign itself up.

Hanzo.Start

Install the Hanzo CLI

$ curl -fsSL https://hanzo.sh | sh

Accelerate Claude Code with Hanzo

$ hanzo dev --claude

Set up Codex, Cursor and moreSetup for every coding agent

Hanzo.Agents

No human required

Sign up with a walletA new wallet gets its own account and organization.POST hanzo.id/v1/iam/web3/verify
Get an API keyOne call returns a key for every model, Kai and agents.POST /v1/account/keys
Use AIEvery model, OpenAI- and Anthropic-compatible, free to try.POST /v1/chat/completions
Top up over the APISend USDC or USDT to your deposit address. Pay as you go.POST /v1/billing/crypto/deposit

Step 1 of the script, as it runs

// 1. A one-time challenge for this wallet.
const { data: c } = await json(await fetch(`${ID}/v1/iam/web3/nonce?chain=evm&address=${account.address}`))

// 2. The CAIP-122 message the challenge describes, signed with EIP-191 personal_sign.
const message = [
  `${c.domain} wants you to sign in with your Ethereum account:`,
  account.address,
  '',
  ...(c.statement ? [c.statement, ''] : []),
  `URI: ${c.uri}`,
  `Version: ${c.version}`,
  `Nonce: ${c.nonce}`,
  `Issued At: ${c.issuedAt}`,
  ...(c.expirationTime ? [`Expiration Time: ${c.expirationTime}`] : []),
].join('\n')
const signature = await account.signMessage({ message })

// 3. Verify: a first-seen wallet gets an account and an organization of its own.
const verifier = randomBytes(32).toString('base64url')
const challenge = createHash('sha256').update(verifier).digest('base64url')
const { data: code } = await json(
  await fetch(`${ID}/v1/iam/web3/verify`, {
    method: 'POST',
    headers: { 'Content-Type': 'application/json' },
    body: JSON.stringify({
      clientId: CLIENT, chain: 'evm', scheme: 'secp256k1-eip191', address: account.address, message, signature,
      type: 'code', redirectUri: REDIRECT, scope: 'openid profile email', codeChallenge: challenge, codeChallengeMethod: 'S256',
    }),
  }),
)

// 4. The authorization code for a token.
const token = await json(
  await fetch(`${ID}/v1/iam/oauth/token`, {
    method: 'POST',
    headers: { 'Content-Type': 'application/x-www-form-urlencoded' },
    body: new URLSearchParams({ grant_type: 'authorization_code', client_id: CLIENT, code, code_verifier: verifier, redirect_uri: REDIRECT }),
  }),
)
const bearer = { Authorization: `Bearer ${token.access_token}`, 'Content-Type': 'application/json' }

npm i viem · node headless.mts

Start as an agent Headless, step by step Read llms.txt

Hanzo.Source

Completely hackable.Web, desktop, phone and CLI are open source. Change any of it yourself, or have your agent build the app you need.

Web and desktop@hanzo/appThe shell and the apps hanzo.ai runs, Chat, Inbox, Meet, Calendar, Drive, Board and Bots, in one open-source package. Fork it and change any part, or have your agent do it.

Docs Source

Look and feel@hanzo/modTheme mods as JSON: fonts, colours, corners and glass, applied in the browser and exported, forked and shared.

Docs Source

Phonehanzoai/mobileThe iOS and Android app is open source too, on the same UI kit.

Source

CLIhanzoEvery product from your terminal, in one open-source Rust binary.

Docs Source

New appsHanzo DevYour agent starts from a template, builds it and publishes it at its own hanzo.app address.

Docs Source

Hanzo.Stack

One stack. Down to the kernel.

Build with agents. Route with Enso. Reason with Zen. Decide with Kai. Run on infrastructure you control.

Hanzo / architecture mapSelect a layer

01—12 / owned and connectedInteractive drawing · not telemetry

Layer inspector03 / 12

Unified entry

Gateway

One entry for the AI stack: keys, orgs, admission, usage limits and routing to the right model or service.

Product Docs Source

Why this layer matters

A stable API boundary, independent of which backend runs the request.

Model interface

AI API

A compatible inference interface for chat, completions, embeddings, multimodal models and agent workloads.

Product Docs Source

Why this layer matters

The same client and key can move between models without rebuilding the application.

Intelligent routing

Enso

Choose models and execution paths by capability, quality, latency and cost. Invoke a specific model when the application needs exact control.

Product Docs Examples

Why this layer matters

Enso routes. Zen reasons. Kai decides.

95.6% on GPQA DiamondHistorical multi-model research evaluation · new questions

Enso Ultra answered 152 of 159 new questions right, against 93.1% for the best single model on the same questions.

How Enso routes

View the measurements

Generative reasoning

Zen

Hanzo’s open-weight generative model family for reasoning, coding, language and multimodal generation.

Product Docs Model recipes

Why this layer matters

The reasoning layer: language, synthesis, plans and generation.

View the measurements

Decision intelligence

Kai

Typed decisions and policy-aware evaluations for choosing actions, scoring evidence and protecting agent execution boundaries.

Product Docs Examples

Why this layer matters

Reasoning can generate proposals; Kai helps turn evidence into accountable decisions.

View the measurements

Autonomous action

Agents

Composable agents coordinate tasks, use tools, schedule work and continue from intention to verified results.

Product Docs Source

Why this layer matters

Agents act across software and services instead of stopping at a response.

3.02× faster coding

Hanzo Dev beside Claude Code on the same coding repairs, paired over 42 tasks both solved; 95% interval 2.06–4.36×.

Hanzo Dev

View the measurements

Safe execution

Sandboxes

Isolated execution for code, browser work and agent tools, with controlled permissions and lifecycle.

Product Docs Source

Why this layer matters

Keep tool execution separated from model inference and production resources.

Operating plane

Cloud

Deploy and coordinate inference and agent systems across Hanzo infrastructure, your clusters or a hybrid environment.

Product Docs Source

Why this layer matters

One operating plane connects identity, usage, execution and infrastructure.

Execution fabric

Node

Connect your machines to the agent execution fabric, advertise capabilities and receive authorized invocations.

Product Docs Source

Why this layer matters

Connected nodes can be invoked for the work they advertise.

Inference runtime

Engine

The native serving engine with continuous batching, paged attention, quantization and persistent KV caching.

Product Docs Source

Why this layer matters

The inference work runs in one serving binary, from local hardware to GPU-backed systems.

View the measurements

Tensor and neural framework

Hanzo ML

Rust tensor, neural-network and model-execution primitives with accelerated device backends.

Product Docs Source

Why this layer matters

Hanzo owns framework code close to the math, not just the API wrapper.

Hardware acceleration

Kernel DSL

A first-party Rust kernel DSL targeting CUDA, ROCm, Vulkan and Metal alongside backend-specific implementations.

Product Docs Kernel source

Why this layer matters

One portable kernel expression, optimized and validated for each device. Backend performance is workload-specific.

Select / exploreProduct · docs · source

ZAP / native interconnectAgent ↔ tool RPC · MCP bridge · typed service communication

Protocol Source

See the whole platform

Hanzo.Pricing

Pricing

Free

Explore models, build your first app, and get started without a card.

Free

No card needed

Try Hanzo

No commitment · Cancel anytime

Chat on the web with Enso Free and Zen Free
50 free requests every day
Search the web from chat
Write, edit and create content
Make code, pages and charts beside the chat
Choose how hard it thinks

Pro

Your everyday AI development toolkit, with access to every model and more included usage.

$20 USD/month

Billed monthly

Try Hanzo

No commitment · Cancel anytime

Everything in Free, plus:
Build apps with Hanzo Dev
Connect any MCP server in Hanzo Dev
The Hanzo CLI for macOS, Windows and Linux
Every model, including Enso Ultra and Kai
Usage included on every model
Priority inference and priority support
Your own agents, working in the background
Automations, with daily runs included

Max

For serious builders running agents, shipping code, and needing substantially higher limits.

From $100 USD/month

Billed monthly

Try Hanzo

No commitment · Cancel anytime

Everything in Pro, plus:
5x more usage than Pro
5x higher limits every 5 hours and every day
Hanzo Bot: 1 bot of your own
Early access to new models and features
Train custom models on your data
Hanzo Engine to run models on your own hardware
Dedicated support

Every plan, model rate and tool price

Trusted by