THE GAP.
Public AI makes ordinary document work faster. Controlled data cannot be placed there without the required authorization and safeguards, so the most sensitive work stays manual.
FOR FEDERAL CONTRACTORS
The deadline moved. The work did not.
Your staff still have proposals, drawings, contract data, and CUI they cannot put into a consumer AI account. Give them a capable machine inside the facility instead.
01CUI STAYS INSIDE
02NO INTERNET CONNECTION
03NO PER-SEAT FEES
FROM $14,500 INSTALLEDFIRST YEAR OF MANAGED CARE INCLUDEDINSTALLED ON SITE ACROSS NORTHERN VIRGINIA AND WASHINGTON DC
01 / WHY LOCAL
THE COMPLIANCE WORK REMAINS.
PUBLIC CLOUD / EXTERNAL
THE BURDEN.
DoD suspended CMMC Phase II after small contractors described the cost as untenable. Phase I self-assessments and existing DFARS safeguarding duties remain. Smaller teams still need a practical place to do the work.
THE EXPOSURE.
Cybersecurity promises in a federal contract are not marketing language. In 2025, MORSECORP agreed to pay $4.6 million to resolve allegations that it failed to meet requirements covering systems used for Army and Air Force work.
PRIMARY SOURCESDOD / CMMC UPDATE DOJ / MORSECORP FAR / CURRENT THRESHOLDS
02 / EXAMPLES
TWO COMMON JOBS
Tuesday, 4:40 PM.
Your proposal is due at five. A contracts manager has a forty-page RFP with a distribution statement on every page. A public summarizer would finish the first pass quickly, but the RFP cannot be uploaded to it.
With a Node on the rack, the manager loads the RFP into a machine on your network. The summary and compliance matrix come back, and the proposal files at 4:58. Nothing leaves the facility.
Thursday, 9:15 AM.
The delivery schedule lives in four spreadsheets that do not agree, and reconciling them takes one morning every week. A staff member asks the Node to build a tool for it. Twenty minutes later, a small program on the shared drive reads all four files, flags every disagreement, and prints the report.
The staff member who requested it is not a programmer.
These are two common jobs: controlled document work and small internal software.
03 / WHAT YOU GET
THE NODE
A sealed machine on your own network, configured and maintained by Splay inside your facility. Staff use it to draft proposals and compliance matrices, write working scripts and small tools, and search the documents you cannot upload: proposals, drawings, contract data.
It needs no cloud account and no internet connection, and nobody pays per seat. We maintain the machine without ever holding your files. The whole setup is written down, step by step; any competent IT shop could run it without us.
04 / WHAT IT DOES
DOCUMENTS, SOFTWARE, AND SEARCH.
A Node runs an AI model inside your building. Staff use it to draft documents, build small tools, and search files that cannot go to a public service.
WORKING SOFTWARE
Staff describe a small tool in plain English. The Node writes the code, and the tool runs on your network.
WEB PAGES
Intranet pages, forms, and working dashboards built from a plain description.
DOCUMENTS
Proposals, compliance matrices, SOPs, contract reviews, letters, and summaries in your templates.
DRAWINGS AND PHOTOS
A vision model can read scanned drawings and photos of parts.
Every Node supports the same types of work. Larger machines handle more users, larger models, or faster answers. The pilot lets you test one with your own work.
05 / THE MACHINES
THE MACHINES
Five standard machines and one custom build. Each runs on your network with no path to the internet. Hardware, capacity, and prices are listed on every card.
63MODEL SCORE / BEST CLOSED MODEL = 100
- MODEL
- MINIMAX-M2.7
- HARDWARE
- DGX SPARK, 128GB
- SERVES
- 1-15 STAFF
$14,500 INSTALLEDYEAR 1 CARE INCLUDED, THEN $600/MO
ASK ABOUT 01 67MODEL SCORE / BEST CLOSED MODEL = 100
- MODEL
- DEEPSEEK V4 FLASH
- HARDWARE
- 2X SPARK, LINKED, 256GB
- SERVES
- 16-30 STAFF, TWO MACHINES
$32,500 INSTALLEDYEAR 1 CARE INCLUDED, THEN $900/MO
ASK ABOUT 02 85MODEL SCORE / BEST CLOSED MODEL = 100
- MODEL
- GLM-5.2
- HARDWARE
- 4X SPARK, LINKED, 512GB
- SERVES
- 1-10 STAFF ON GLM-5.2, OR 60 SPLIT
$59,500 INSTALLEDYEAR 1 CARE INCLUDED, THEN $1,200/MO
ASK ABOUT 03 73MODEL SCORE / BEST CLOSED MODEL = 100
- MODEL
- MINIMAX-M3
- HARDWARE
- 4X RTX PRO 6000, 384GB
- SERVES
- 15-75 STAFF, FAST
FROM $129,500 INSTALLEDYEAR 1 CARE INCLUDED, THEN $1,500/MO
ASK ABOUT 04 85MODEL SCORE / BEST CLOSED MODEL = 100
- MODEL
- GLM-5.2, BEST OPEN MODEL
- HARDWARE
- DGX STATION GB300, 748GB
- SERVES
- 15-75 STAFF
FROM $195,000 INSTALLEDYEAR 1 CARE INCLUDED, THEN $2,500/MO
ASK ABOUT 09 If the listed machines do not fit your workload, Splay can build one to order. Memory, speed, and models are set after a call and quoted in writing.
BY INQUIRYSIZED TO YOUR WORK / SCOPED ON A CALL
ASK ABOUT NODE X 01START
NODE 01 serves 1-15 staff.
02JOIN
A second Spark and one cable turn NODE 01 into NODE 02.
03GROW
Two more make NODE 03. Nothing is replaced or migrated.
04CREDIT
What you paid counts in full from NODE 01 through NODE 03.
NODE 03 is built for more users. NODE 04 is built for faster answers. NODE 04 uses four RTX PRO 6000 cards, each with 1,792 GB/s of memory bandwidth. One Spark has 273 GB/s. Both can run GLM-5.2.
You can change the model without replacing the machine. Offices that require American-made models can run gpt-oss, Llama, or Nemotron. No model can send anything out. At $14,500, NODE 01 falls below the current $15,000 federal micro-purchase threshold; purchaser authority and agency procedures still apply.
MODEL LEDGER / JULY 2026
Pick a model. See where it lands.
Published open models against the machines. The verdict is arithmetic over each model's published size and each machine's installed memory, nothing else. The flag is the developer's home country.
NODE 01 / 128GBNO FIT
NODE 02 / 256GBNO FIT
NODE 03 / 512GBQ4
NODE 04 / 384GBQ3
NODE 09 / 748GBQ4
The estimate covers the model itself plus 5% overhead; working memory is extra. Sizes are the published figures.
Model score: best closed model = 100 on the Artificial Analysis index, July 2026. Raw scores 38, 40, 51, 44, 51 against the best closed model's 60. Model fit includes 5% overhead; working memory is extra. NODE 04 and 09 are built to order: 50% deposit, quotes hold 30 days. Prices subject to change.
06 / COST
COST COMPARISON
THE QUESTIONSPLAY NODEGOV-CLOUD AIPUBLIC AI TOOLS
PERMITTED FOR CUIYESYES, AFTER MIGRATIONNO
YOUR DATA LIVESIN YOUR BUILDINGMICROSOFT GOV CLOUDON VENDOR SERVERS
SETUPFROM $14,500MIGRATION + CONFIGURATIONNONE
RUNNING COST$0 YR 1, THEN $600/MOTENANT + LICENSES$20-30/SEAT/MO
COST TO ADD A SEAT$0ANOTHER LICENSEANOTHER SEAT
Public tools remain fine for public work. Gov-cloud migration and licensing vary by tenant and are quoted by the provider. Hardware alone does not certify you.
STAFF / 15$28,900 / NODE 01 / 3 YEARS
GOV-CLOUDMIGRATION + TENANT + LICENSES
MODEL SCORE / NODE 0163 / SELECTABLE
07 / PRICING
FULL PRICE LIST
These are Splay's current published prices.
THE CLEAN ROOM
A CUI enclave, designed and built inside your facility: its own walled-off network, a bay for your Node, screens that can view but never copy, and a documentation pack for your assessor.
$25,000 TO 10 SEATS$45,000 TO 25 SEATSFixed price; the machine is priced separately. You own it, and your IT runs it with our manual.
SERVICES
- PILOT, 30 DAYS
- $2,000*
- READINESS SURVEY
- $2,400*
- EXEC BRIEFING
- $1,500
- COURSE, HALF DAY
- $2,400
- COURSE, FULL DAY
- $5,500
- ONBOARDING
- FROM $4,500
- CUSTOM MODEL
- FROM $12,500
* Credited in full against any machine. Courses use your own examples. Custom models are tuned on your hardware. Extra on-site work is $2,400 a day.
ON RETAINER
- MANAGED CARE
- FROM $600/MO
- MODEL DROP
- $250/QTR
- AI OFFICER
- FROM $2,500/MO
Model drop: each quarter's best open model, installed and briefed on-site. Included with NODE 04/09 care. AI officer: a standing operator for prompts, workflows, and small tools.
08 / FROM CALL TO INSTALL
HOW IT WORKS
THE CALL
A 20-minute call about your network, staff, and work.
THE SURVEY
A readiness survey at your facility. $2,400, credited in full against any machine.
THE PILOT
A pilot machine on your bench for 30 days. $2,000, credited in full. Then you decide.
A fleet of four pilot machines exists. Every machine is a copy of one documented setup, with a step-by-step manual; any competent IT shop could run it without us.
09 / FAQ
QUESTIONS
Is this CMMC certification?
No, and no machine is. Hardware alone does not certify you. The Node gives your staff a permitted place to work with controlled data, and the clean room engagement includes a documentation pack for your assessor.
The strongest open models are Chinese. Can we run American models?
Yes. You can change the model without replacing the machine. Offices that require American-made models can run gpt-oss, Llama, or Nemotron. NVIDIA's Nemotron 3 Ultra scores 38 on the same public index used on this site, level with NODE 01's standard model. The machine is sealed either way: no model can send anything out.
Can it write software?
Yes. Staff describe a tool in plain English and the machine writes the code. It works best for small internal tools and scripts. Large production systems still need a software team. The pilot lets you test it with your own work.
Who runs it after the install?
Your IT does, with our manual. Every machine is a copy of one documented setup. If you would rather we watch it, managed care starts at $600 a month and year one is included with every machine.
What happens if Splay disappears?
You own the machine outright, and the whole setup is written down, step by step. Any competent IT shop could run it without us.
Can we try it before we buy it?
A pilot machine sits on your bench for 30 days for $2,000, credited in full against any machine. So is the $2,400 readiness survey.
What does it cost to add a seat?
$0. The machine is the cost. NODE 01 serves 1 to 15 staff; when you outgrow a machine you buy the next machine, not the next license.
How many people can use it at once?
Each person uses the Node over your network. What limits a machine is how many answers it writes at the same time. NODE 01 is bench-measured at ten simultaneous conversations. Larger machines serve more people, and Splay sets the working memory during installation so the machine stays responsive.
What happens when we outgrow it?
You add machines. A second Spark turns NODE 01 into NODE 02; four Sparks make NODE 03. What you paid is credited in full from NODE 01 through NODE 03. Credits toward NODE 04 and NODE 09 are set by quote.
Do you see our data?
No. The machine has no path out to the internet, and your files stay in your building. Custom models are tuned and tested on your hardware, and support visits happen on-site, at the rate on the card.
Why are your prices public?
So buyers can compare the machines before a call. Model scores come from a public index, and the raw scores are printed on the page. Every price carries its edition date: July 22, 2026.
10 / CONTACT
TALK TO SPLAY
Splay is based in Stafford, Virginia, and installs on site across Northern Virginia—including Arlington, Alexandria, Fairfax, Loudoun, and Prince William—and Washington, DC. Start with a 20-minute call.