SPLAYCapacity upgrades
Compatible systems can be linked to upgrade from NODE 01 through NODE 03. Larger systems and expansion plans are scoped during consultation.
MACHINES / PRE-ORDER
Local AI
configurations.
SPLAY installs private AI systems on new commercial NVIDIA-based hardware in your building. You own the installed system.
- AVAILABILITY
- By consultation only
- LAUNCH CARE
- 90 days included
- ONGOING SUPPORT
- Optional
PRE-ORDER ONLY / GENERAL SALES NOT YET OPEN
COMPARISON
Compare configurations
Compare memory, baseline models, and intended workloads. These are reference configurations; Splay confirms your selection and final requirements during consultation.
| SYSTEM | MEMORY | BASELINE MODEL | INTENDED USE | Inquiry |
|---|---|---|---|---|
| 01NODE 01 | 128 GB | SPLAY 35B A3B | 1-15 STAFF | |
| 02NODE 02 | 256 GB | SPLAY FLASH | 16-30 STAFF, TWO MACHINES | |
| 03NODE 03 | 512 GB | SPLAY HEAVY | TEAM CAPACITY CONFIRMED DURING CONSULTATION | |
| 04NODE 04 | 384 GB | SPLAY FASH FAST | 15-75 STAFF | |
| 05NODE 09 | 748 GB | SPLAY ULTRA | 15-75 STAFF | |
| 06NODE X | Specified after consultation | Selected after consultation | CUSTOM |
Larger models and more simultaneous work require more capacity. Staff counts are sizing estimates, not guarantees of simultaneous full-speed use.
SPECIFICATION SHEETS
Configuration details
Expand a configuration to view its hardware, intended workload, and optional support.
01NODE 01
128 GBBy consultation only
MEMORY CAPACITYGB
128GBYour quote specifies the physical configuration.
- HARDWARE
- NVIDIA-BASED DESK SYSTEM, 128GB
- BASELINE MODEL
- SPLAY 35B A3B / BASELINE
- INTENDED USE
- 1-15 STAFF
- PREVIOUS MODEL BENCH TEST
- 10 simultaneous conversations with LAGUNA S 2.1
- MANAGED CARE
- Optional after Launch Care. By consultation only.
- INTEGRATION
- SPLAY / Stafford, Virginia
02NODE 02
256 GBBy consultation only
MEMORY CAPACITYGB
256GBYour quote specifies the physical configuration.
- HARDWARE
- LINKED NVIDIA-BASED SYSTEMS, 256GB
- BASELINE MODEL
- SPLAY FLASH
- INTENDED USE
- 16-30 STAFF, TWO MACHINES
- MANAGED CARE
- Optional after Launch Care. By consultation only.
- INTEGRATION
- SPLAY / Stafford, Virginia
03NODE 03
512 GBBy consultation only
MEMORY CAPACITYGB
512GBYour quote specifies the physical configuration.
- HARDWARE
- NVIDIA-BASED CLUSTER, 512GB
- BASELINE MODEL
- SPLAY HEAVY
- INTENDED USE
- TEAM CAPACITY CONFIRMED DURING CONSULTATION
- MANAGED CARE
- Optional after Launch Care. By consultation only.
- INTEGRATION
- SPLAY / Stafford, Virginia
04NODE 04
384 GBBy consultation only
MEMORY CAPACITYGB
384GBYour quote specifies the physical configuration.
- HARDWARE
- NVIDIA PROFESSIONAL GPU SYSTEM, 384GB
- BASELINE MODEL
- SPLAY FASH FAST
- INTENDED USE
- 15-75 STAFF
- MANAGED CARE
- Optional after Launch Care. By consultation only.
- INTEGRATION
- SPLAY / Stafford, Virginia
05NODE 09
748 GBBy consultation only
MEMORY CAPACITYGB
748GBYour quote specifies the physical configuration.
- HARDWARE
- NVIDIA AI WORKSTATION, 748GB
- BASELINE MODEL
- SPLAY ULTRA
- INTENDED USE
- 15-75 STAFF
- MANAGED CARE
- Optional after Launch Care. By consultation only.
- INTEGRATION
- SPLAY / Stafford, Virginia
06NODE X
CUSTOMBy consultation only
Custom configurations are available when the standard machines do not meet your requirements. Memory, speed, models, and care are specified during consultation and recorded in your written proposal.
DISCUSS A CONFIGURATIONDELIVERY STANDARD
Included with installation
Each installation includes hardware, SplayOS, a validated model, local interface, outbound network controls, recovery testing, documentation, on-site installation, and 90 days of Launch Care.
Delivery and payment terms
Hardware warranty, delivery timing, payment, and acceptance terms are agreed during consultation and recorded in your written proposal. NODE 04 and NODE 09 are built to order.
Hardware sourcing and documentation
SPLAY sources new commercial hardware through established manufacturers and U.S. retail or distribution channels. No grey-market, used, or refurbished components are used unless the buyer approves them in writing.
Every delivery records the platform, configuration, serial numbers, and firmware state. Buyers with contract-specific restrictions can approve the hardware before SPLAY purchases it.
Technical appendix / model memory estimates
MODEL COMPARISON / JULY 2026
Model memory estimates
The current SPLAY baseline models are listed in the catalog above. This appendix compares published open models to estimate whether they fit in each machine's memory. The calculation uses published model sizes and installed memory; it does not measure speed or full workload capacity. Flags indicate the developer's home country.
NODE 01 / 128GBQ4
NODE 02 / 256GBFP8
NODE 03 / 512GBFP8
NODE 04 / 384GBFP8
NODE 09 / 748GBFP8
NODE 01 + SPLAY BUILDHISTORICAL TEST / LAGUNA S 2.1
QUALITY CHECKS9/9
SINGLE STREAM26.5 TOK/S
5 + 10 STREAMS56.0 TOK/S
TESTED CONTEXT4 × 65,536
These measurements used the previous Laguna S 2.1 baseline, not SPLAY 35B A3B. SPLAY verifies the selected configuration before delivery. TOK/S means tokens per second.
FP8, Q4, and Q3 indicate 8-bit, 4-bit, and 3-bit model precision. The estimate reserves 5% of installed memory and allows 5% overhead for the model. Working memory is extra. Model sizes are published figures.
QUESTIONS
Frequently asked questions
Who operates the machine after installation?
Your IT staff operate the system using the supplied documentation. Every machine includes 90 days of Launch Care for handoff, initial tuning, staff setup, and issue resolution. Managed Care is optional afterward for scheduled health checks, priority support, configuration maintenance, and planned on-site service. It does not give SPLAY unattended remote access to your files or Node.
Can we operate the system without SPLAY?
You own the machine outright. The setup and operating procedures are documented so your IT staff or another service provider can operate it without SPLAY.
What happens if the machine fails?
SplayOS saves users, chats, documents, settings, and the system credentials needed for recovery in a single snapshot. The complete snapshot, reset, provisioning, and restore process has passed testing on NODE 01 hardware. A current customer-held snapshot can be restored to a replacement machine. SPLAY keeps no copy of it.
Buying guides
Plan your local AI system
Compare deployment options, prepare hardware requirements, and decide which files the model can use.
Local AI vs. cloud AI for business
Compare local and cloud AI by file handling, hardware ownership, internet access, maintenance, and workload. A guide for businesses choosing where AI runs.
How to size an on-premises AI system
Choose local AI hardware around the model, document length, simultaneous users, and required capabilities. See what to prepare for a SPLAY consultation.
Private AI search for business documents
Plan local AI search across authorized business files. Learn what to check for document preparation, staff access, source references, and answer quality.
Contact
Discuss your installation
Machines and services are by consultation only. Discuss your workload, network, and preferred configuration with Splay.