01
Private host
Only your workloads run on it. Your references, assets and review history stay with the deployment.
Dedicated deployment
Dedicated MiniMax H3 deployment for enterprise
Nuva Lab runs MiniMax H3 on a private host that serves only your company, tuned for speed. You get the full concurrency of the GPUs, a model customized on your work, agents that run production, and a system that improves with your review data.
What you get
A production system, not an endpoint
02
Full concurrency
Every GPU serves your jobs. Large batches do not wait in a shared queue.
03
Your model
We fine-tune and post-train H3 on your approved work and run your LoRAs. Fine-tuning and LoRA.
04
Speed
Eight RTX PRO 6000 GPUs render about 4,000 ten-second 768p clips with audio per day at full use. We tune the model and the serving stack for your workload. B200, B300 and GB200 hosts are available on request. Inference speed data.
05
Agents
The Creative Agent runs briefs, batches, retries and checks. The FDE Agent helps deploy and maintain the host. How the agents work.
06
Data flywheel
Your approvals and rejects guide the next tuning round. The loop runs inside your deployment and uses your data only.
Cost
One fixed cost: about $0.014 per second at full use
Illustrative GPU cost: eight RTX PRO 6000 GPUs at $3 per GPU-hour cost $576 per day. At full use they render about 4,000 ten-second 768p clips with audio per day, or about $0.0144 per second of video.
This is GPU cost only. It excludes service fees, storage and network. The MiniMax H3 API list price is $0.08 per second at 768P. Cost guide and calculator.
B200, B300 and GB200 hosts are available on request. Contact us for a quote.
US companies
Authorized for US companies
The MiniMax H3 Community License excludes the US by default. Nuva Lab holds a MiniMax authorization for H3 deployments that serve US companies.
Use of H3 and its derivatives follows the MiniMax H3 license.
Workloads
Built for production video
| Workload | What the deployment holds consistent |
|---|---|
| Performance ads | Products, logos, packaging and brand colors across many variants |
| Game marketing | Characters, art style and gameplay look across creative tests |
| Short drama | Characters, wardrobe and setting across shots and episodes |
How it starts
From first call to production
| Step | What happens |
|---|---|
| 1. Scope | We review your volume, clip formats and review criteria. |
| 2. Pilot | We deploy the model and test it on your briefs. |
| 3. Customize | We tune on your approved work and measure the acceptance rate. |
| 4. Production | Agents run your workflows. We keep the deployment upgraded. |
FAQ
Questions
What is a dedicated MiniMax H3 deployment?
A private host that runs MiniMax H3 for one company only, tuned for speed. It gives full concurrency, runs your customized weights and keeps your data inside the deployment.
How is it different from the MiniMax H3 API?
The API is a stateless, shared service that charges per generated second. A dedicated host is private, runs your fine-tuned model and improves with your review data.
Can US companies use MiniMax H3?
The MiniMax H3 Community License excludes the US by default. Nuva Lab holds a MiniMax authorization for H3 deployments that serve US companies.
Is my data used to improve models for other customers?
No. The feedback loop runs inside your deployment and uses your data only.
Which GPUs does the host use?
The standard host has eight NVIDIA RTX PRO 6000 GPUs. The speeds and prices on this site are for that host. For B200, B300 or GB200 hosts, contact us.
Sources and further reading
Read next
Get a dedicated MiniMax H3 deployment
Tell us your volume, formats and review process. We'll size a private host for it.
Get started · 1 min