Hanzo AI

Frontier AI,
one line away.

Chat with frontier models, run Claude Code or Hanzo Dev on Hanzo, and ask Kai for decisions, on one account. Sign up free with Google or GitHub.

or with email

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

Hanzo.Start

Install the Hanzo CLI

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

Run Claude Code on Hanzo

$ hanzo login$ hanzo dev --claude --cloud

Or start your own Claude Code on Hanzo

$ hanzo login$ env -u ANTHROPIC_API_KEY claude --settings '{"apiKeyHelper":"hanzo auth token","env":{"ANTHROPIC_BASE_URL":"https://api.hanzo.ai"}}'

Paste into your coding agent

Read https://hanzo.ai/skill.md and follow it to set up a Hanzo company in my project: register or sign in with `hanzo login`, pick a starter with `hanzo template list`, clone the repository `hanzo template get <slug>` names in its `source`, then tell me to run `hanzo dev` in it to start building. Check you are in the project directory first — if not, ask me where it is.

Glider 1.0

Coding.Tasks

3.02× faster on the same coding tasks.

Hanzo Dev and Claude Code were handed the same 44 coding tasks (single-fault repairs), one try each. Each clock replays that side's median time a task, at the speed it ran.

Claude Code

done

median time a task

9.52 s

model calls a task3.8

replay of run coding.repair-held-devfix-claude-code

Hanzo Dev

done

median time a task

2.29 s

model calls a task1.0

replay of run coding.repair-held-devfix-hanzo-dev-dc

3.02× faster over the 42 tasks both solved, 95% interval 2.06–4.36×.By how Hanzo Dev settled a task:

Settled by search, no model call, 24 tasks8.22× · 95% 6.86–9.82×
The model’s first edit, 4 tasks1.31× · 95% 1.14–1.48×

Every coding task and both runs →

GPQA.Diamond

Frontier answers.

GPQA Diamond is hard science questions, the same for every model. Each bar is a research run.

Reading Enso's research runs from api.hanzo.ai

Price.List

Price per million tokens.

Input and output, in dollars, as our price list states them.

Reading the price list from api.hanzo.ai

Every model and rate →

Clock.Speed

Zen 6, on our own hardware.

Measured by Hanzo.

Reading the decision timings from api.hanzo.ai

Zen6.Prefill

~2,400 tok/s

Zen 6 reads ~2,400 prompt tokens a second on NVIDIA Grace-Blackwell, NVFP4 weights.Decode ~43 tokens a second. Measured by Hanzo on Sep 23, 2026.Zen 6 on three machines →

Hanzo.Apps

One account. Every tool you build with.

How do black holes form?

A massive star exhausts its fuel and gravity collapses the core into a singularity.

What causes aurora borealis?

Solar particles hit atmospheric gases near the poles, exciting them to glow.

ChatExplore

projects/main

14.40%

▾ thought for 4.1s

◆ Editsrc/middleware/auth.ts

42export async function handler(req) {

43 const token = extractBearer(req);

44 if (!token) return unauthorized();

47 const session = await getSession(req);

48 req.user = payload;

› Add rate limiting to all API routes.

DevExplore

I'll match every charge to a receipt and file the report. Anything that doesn't line up, I ask instead of guessing.

Created routine · Month-end close

Thinking...

BotExplore

StudioExplore

RealtimeExplore

Research.Log

The latest results, with every task.

Each result is read from research runs on public questions, the same for every model.

Research.DecisionSuite

Kai

Reading the Decision Suite's runs from api.hanzo.ai

Reading Enso's research runs from api.hanzo.ai

Transparent valueNo hidden fees

Build what’s next

You pay exactly the price shown. Start free; paid plans include usage on every model.

Reading the plans from api.hanzo.ai

Paid plans include usage on every model. Past it, switch to Enso Free or keep going with prepaid credit.

Upgrade Plan →

FAQ

What is Hanzo and how does it work?

Hanzo is an AI cloud on one account. Chat with frontier models, build and ship with Hanzo Dev, and call the same models from one API: Enso, our frontier model, and Kai, which makes a decision in one call. How the platform fits together.

Can I keep using Claude Code?

Yes. Install the Hanzo CLI and sign in, then run Claude Code with its model calls going through api.hanzo.ai, or point your own Claude Code install at Hanzo. The commands are at the top of this page, ready to copy.

Where do the numbers on this page come from?

Each one is read when the page opens: the race and the boards from Hanzo's research runs, the prices from our price list, and the plans from our plan list. A figure whose source does not answer is left out, never replaced with an older one. Every board and its runs.

How much does it cost to use?

Free needs no card. Paid plans are sold by the month for one person and by the member for a team, and every paid plan can be cancelled any time. Enterprise is priced with our team. See every plan and model rate.

Can Hanzo run in our own cloud or on our own GPUs?

Yes, with Enterprise: in your cloud, on your own servers or fully offline, with dedicated compute and your own models. Hanzo Enterprise.

Frontier AI,one line away.

3.02× faster on the same coding tasks.

Frontier answers.

Price per million tokens.

Zen 6, on our own hardware.

One account. Every tool you build with.

The latest results, with every task.

Build what’s next

FAQ

3.02× faster on the same coding tasks.

Frontier answers.

Price per million tokens.

Zen 6, on our own hardware.

One account. Every tool you build with.

The latest results, with every task.

Build what’s next

FAQ

Frontier AI,
one line away.