Banks
Regulator technology-risk and model-risk rules, met by default — every prompt, model, and log stays inside the perimeter, ready for the next exam.

A sealed, on-premise agent factory that runs thousands of AI agents entirely inside your walls. No public cloud. No external model calls. No token meter.

Regulator technology-risk and model-risk rules, met by default — every prompt, model, and log stays inside the perimeter, ready for the next exam.

Claims files, actuarial models, and customer records reasoned over without a single external call — with the decision logs the NAIC bulletin expects.

Citizen data stays sovereign — deployed inside national borders, operated by your own administrators, with audit trails that answer to public-records law.

Built for air-gapped and classified networks: complete traceability, updates by signed physical media, and a single accountable supply chain.

Patient records and clinical notes processed where they live — ambient documentation, coding, and triage with PHI never leaving the building.

Privileged work product stays privileged — review, drafting, and matter search on-premise, satisfying the strictest outside-counsel guidelines.

Process recipes, CAD, and yield data are trade secrets — troubleshooting, quality, and maintenance AI that runs inside the OT boundary with zero external calls.

Biometric gates, cashless payments, and crowd analytics processed on venue hardware — real-time AI with biometric data kept on-site.
— and any country or company with data-localization requirements.
The on-prem advantage

Data sovereignty
Models, prompts, logs, embeddings — everything lives on drives you own, inside your borders, under your law. No cloud region, no third country, no fine print.
Agent Registry
Every agent on the box — versioned and governed.
The complete Lyzr stack
Lyzr Studio, LyzrGPT with Enterprise Search, and Agentic OS ship on the box — build agents, work with them and run whole departments the day it powers on.
Enterprise security
End-to-end encryption and tamper-evident logs that satisfy examiners — not just checklists.
Role granted — Analyst
Approval — 2 of 2 signed
Enterprise controls
SSO, role-based access, approvals and a tamper-evident audit log — governed from one pane.
Regulatory compliance
Nothing leaves the building — so GDPR, HIPAA and data-localization reviews get short.
Follow the appliance from bare silicon to running software — the hardware, then every app your teams will live in.
01 · The hardware
Built for regulated industries — with every hardware element it needs already inside: enterprise GPUs, memory, storage, redundant power, silent cooling. All of it chosen, configured and taken care of before it reaches you.
02 · Lyzr Studio
Build, govern and manage agents on the same machine that runs them — visual Superflows, knowledge bases, a simulation engine and full traceability.
03 · LyzrGPT & Enterprise Search
Any model, brainstorming, artifacts, custom agents — with Enterprise Search built in, so answers come from your own data with permissions intact.
04 · Agentic OS
Packaged workbenches your teams build and deploy — finance, HR, marketing, sales — all running inside your walls.
THE ON-PREM AI AGENT FACTORY
One sealed machine that runs your agents, models and enterprise search entirely inside your walls. No cloud. No meter.

The enterprise inference node. Eight Blackwell GPUs serving open models to your entire company from a single 4U chassis.
Year one
$750,000
Year two onwards
$300,000 / yr
Includes the Lyzr Studio annual subscription.
Serves models from
Fully refundable until shipped
OPTIMUS MINI · TECHNICAL SPECIFICATIONS
OPTIMUS MINI · MODEL MENU
US and EU models · resident on-device.
GPT-OSS-120B
OpenAI · US
Reasoning model for agents and tool calls — faster and lighter per conversation
Mistral Small 4
Mistral · France
Comprehensive document reading and vision · enables EU languages
Gemma 4 26B
Google · US
Smaller, faster tasks · vision capabilities
Whisper large-v3
OpenAI · US
The speech-to-text model
WHY LYZR OPTIMUS
Six reasons your intelligence stays inside your walls — hardened, sovereign, and entirely your property.
Hardened appliance with secure boot, TPM attestation, and a single controlled network door. Wireless disabled by default.
Runs fully offline inside your perimeter. Zero public-cloud dependency, zero external model calls, zero API exposure.
Fixed infrastructure and predictable spend. No token billing as you scale.
Your agents, prompts, and fine-tunes are your property.
Encrypted at rest with self-encrypting drives, redundant storage, remote wipe, and a tamper-evident audit trail.
Data never leaves your environment. Nothing is logged, trained on, or seen by a third party. Local inference only.
You choose how much health data leaves, from full reporting down to a fully dark site.
The common failure parts—drives and power supplies—swap from the front of the machine, with replacements shipped in advance.
Your data never travels during a repair, and failed drives stay with you or are erased beyond recovery.
Updates arrive as signed bundles, applied with automatic rollback, on physical media for air-gapped sites.
Optimus Mini clusters, so growth means adding a machine.