> ## Documentation Index
> Fetch the complete documentation index at: https://inference.docs.tensormesh.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Reserved Deployments

> Dedicated GPU clusters for high-volume production workloads and enterprise-grade inference.

Reserved Deployments provide dedicated GPU clusters tailored to your specific infrastructure needs — guaranteed capacity, consistent performance, and no resource contention.

Navigate to **Deploy → Reserved** to submit a request.

<img src="https://mintcdn.com/tensormesh-lmignite/FHmfmMW8HKJdXOEW/images/reserved-light.png?fit=max&auto=format&n=FHmfmMW8HKJdXOEW&q=85&s=859778c6b0c0f8156c78211356632ba4" className="hidden dark:block" alt="Reserved" width="2904" height="1658" data-path="images/reserved-light.png" />

<img src="https://mintcdn.com/tensormesh-lmignite/FHmfmMW8HKJdXOEW/images/reserved-dark.png?fit=max&auto=format&n=FHmfmMW8HKJdXOEW&q=85&s=c4198eb7fdbd06ee8f70470ef9a91bd7" className="block dark:hidden" alt="Reserved" width="2910" height="1662" data-path="images/reserved-dark.png" />

***

## When To Use Reserved

<CardGroup cols={2}>
  <Card title="High-Volume Production" icon="chart-line">
    Workloads that require consistent throughput at a scale where serverless costs exceed a flat cluster rate.
  </Card>

  <Card title="Latency SLAs" icon="bolt">
    Applications with strict latency requirements that need dedicated, non-shared GPU resources.
  </Card>

  <Card title="Enterprise Compliance" icon="shield">
    Deployments that require data isolation, custom networking, or specific compliance guarantees.
  </Card>

  <Card title="Tailored Pricing" icon="dollar-sign">
    A flat cluster rate replaces variable per-token billing — easier to budget at scale and priced to your specific workload and capacity requirements.
  </Card>
</CardGroup>

***

## Pricing

Reserved deployments start from **\$250k / year**. A 20% deposit is required upfront to reserve your nodes, and pricing is locked for the full contract term.

<CardGroup cols={3}>
  <Card title="Dedicated Nodes" icon="shield-check">
    Nodes are allocated exclusively to your account and not shared with other tenants.
  </Card>

  <Card title="Rate Lock" icon="lock">
    Pricing is fixed for the duration of your contract term.
  </Card>

  <Card title="SLA-Backed" icon="badge-check">
    Uptime guarantee backed by a formal SLA agreement.
  </Card>
</CardGroup>

***

## Requesting a Cluster

Submit a request through the form at **Deploy → Reserved**.

**Contact Info**

**Full Name** — Your name (pre-filled from your profile)\
**Work Email** — Your business email address\
**Company** — Your organization name\
**Job Title** — Your role (optional)

**Cluster Requirements**

**Preferred GPU** — Choose from NVIDIA H200 or NVIDIA B200\
**Total GPUs Needed** — Capacity is reserved in full nodes of 8 GPUs each; select 8 (1 node), 16 (2 nodes), or 24 (3 nodes)\
**Contract Length** — Choose a 1-year (12 months) or 2-year (24 months) term; pricing is locked for the full duration\
**Workload / Use Case** — Select the primary workload type: Inference / Serving, RAG / Retrieval-Augmented Apps, Agentic / Tool-Using Systems, or Other

<Info>
  Use the **Networking / Storage** field to describe bandwidth targets, storage needs, or interconnect preferences. Use **Other Notes** for compliance requirements, SLA expectations, or anything else the team should know.
</Info>

***

## What Happens Next

<Steps>
  <Step title="Submit Your Request">
    Fill out the form with your GPU and workload requirements and click **Submit Request**.
  </Step>

  <Step title="We Reach Out">
    Our team reviews your needs and follows up within 1 business day with a custom proposal.
  </Step>

  <Step title="Nodes Provisioned">
    Once the contract is signed and deposit received, we spin up your dedicated cluster.
  </Step>
</Steps>

***

## Not Ready for Reserved?

Start with [Serverless Inference](/serverless-inference) — instant access, pay-per-token, no setup required. Serverless is suitable for most development and production API workloads. Move to reserved when volume or latency requirements exceed what serverless offers.

<Tip>
  You can also reach us directly via **Management → Contact Us** to discuss your capacity needs before submitting a formal request.
</Tip>