Hanzo Enterprise
Give every team AI.
Run it where you choose.
Models, agents and one shared workspace for everyone in your company, in your cloud, on your servers or fully offline. Hands-on onboarding, and engineering support around the clock.
What you get
What Enterprise includes.
Your own infrastructure, dedicated compute, and engineering support around the clock.
Your infrastructure
Run Hanzo in your cloud, on your own servers, or fully offline.
Dedicated compute
Compute set aside for your organization alone.
Your own models
Train and host your own models, on your own data.
Commitments in writing
An uptime guarantee and a security review, both in your contract.
24/7 engineering support
Engineering support around the clock, and hands-on onboarding to get your teams started.
Your whole company
Unlimited members, single sign-on, roles and an audit log.
Deployment
Run it where your data lives.
One platform, four places to run it. Choose one, or start in our cloud and move when you are ready.
Hanzo Cloud
We run it for youStart today. On Hanzo Base, your organization's data is its own database.
Your cloud
Inside your VPCHanzo runs in your own cloud account, on your network.
Your servers
On premisesOn hardware in your data center, beside the systems you already run.
Air-gapped
Fully offlineSealed inside your network, with models served on machines you control.
Savings
What the same tokens cost.
Put in what your teams use today. The page prices it at the providers’ list rates, on Hanzo Cloud, and on GPUs of your own, every rate read from our price list as you look.
Your own GPUs
Run the models on machines you own.
Zen 6 runs on one DGX Spark. Zen 6 Max wants a rack of Blackwell servers. What we measured, and what a rack is made of.
Zen 6 on one DGX Spark, as Hanzo measured it
| Prompt | Prefill time | Prefill rate |
|---|---|---|
| 512 tokens | 1.9 s | 270 tokens a second |
| 2,048 tokens | 0.8 s | 2,658 tokens a second |
| 8,192 tokens | 4.5 s | 1,835 tokens a second |
| 16,384 tokens | 9.6 s | 1,701 tokens a second |
| 32,768 tokens | 23.2 s | 1,414 tokens a second |
A cold prompt of about that many tokens on a quiet box, measured Sep 20, 2026: NVIDIA DGX Spark (GB10), 121 GiB unified memory, shared with the benchmark's other jobs; zen6, NVFP4 with FP8 attention, weights 21.6 GiB (serve-spark-zen6.sh). NVIDIA DGX Spark: GB10, 273 GB/s memory bandwidth.
We recommend at least four 8-GPU NVIDIA Blackwell servers, B300 or B200, for Zen 6 Max on your own premises. Routing cuts the average inference work a request needs. It does not cut the memory needed to keep every specialist ready.
DGX B300
DGX B200
Four B300 servers hold 9,216 GB of GPU memory; four B200 servers hold 5,760 GB.
Planning range
Qualification. Four 8-GPU servers is where a qualification starts.
Isolated production pools. One pool for each specialist, so a busy class of work never takes memory from another.
Redundant pools. Two independent replicas of those pools across failure domains.
Around the servers
NVLink and NVSwitch inside each server.
400 or 800 Gb/s RDMA wherever a specialist spans servers.
About 31 TB of NVMe in each server (8 x 3.84 TB).
Separate sandbox workers for coding agents.
A high-availability control cluster.
Final capacity and performance commitments come from qualifying your workload: context length, concurrency, routing mix, latency, and failover.
Proof
Results on the record.
Our models' scores come from research runs anyone can open. Customer results come from the case studies of launches Hanzo ran.
in bikes sold in 30 days
Damon Motorcycles · CES launch and preorders
raised in preorders
KANOA · Crowdfunding and preorders
people watched TrillerFest
Triller · Growth marketing
raised on Indiegogo
SKULLY · Crowdfunding
CSPR peak market value
Casper Labs · Network launch
email signups
Bellabeat · Preorders
Partners
Work with Hanzo.
Refer customers, resell Hanzo under your own brand, or have Hanzo’s agency build it with you.
Refer customers
Bring organizations to Hanzo and earn a share of Hanzo’s margin on what they spend, every period they spend. The customer’s price does not change.
Have Hanzo build it
Hanzo’s agency ships the work with you, from one workstream to an embedded team.
Hanzo runs in your cloud and on these. They are the clouds and the silicon Hanzo deploys on, not partners or customers.
Security
Ready for your security review.
A security review is part of every Enterprise contract. We answer your questionnaire against the code, and the engineers who wrote it walk your reviewers through it.
Single sign-on
Sign in through your own OpenID Connect provider. SCIM 2.0 creates, updates and removes accounts from your directory.
Your data, apart
On Hanzo Base, each organization's data is its own database file, kept apart from every other organization. Identity comes only from a verified token.
Audit log
Every request is recorded: who acted, from where, what they called and the status it returned.
Roles and second factors
Grant access by organization, workspace and project. Every account can add an authenticator app, or a code by SMS or email.
Questions
How is Enterprise priced?
Where do the savings figures come from?
Can Hanzo run fully offline?
How is our data used?
Which identity providers work?
Can we start on Team and move up?
Tell us what you want to build.
Share a few details about your company. We come back with a plan: where Hanzo runs, what it costs and how you start.
Or email [email protected]