Sovereign GaiXer explained
The AI-OS that makes your AI server ready for generative AI from day one
PGX can run meeting-minutes and presentation apps as well as chat because the Sovereign GaiXer AI-OS and the GaiXer App Store sit on top of the AI server hardware. Buying the same PGX alone does not provide business-ready generative AI without that AI-OS. Here is how it works, using the smartphone you use every day as an analogy.
Key points
Like a smartphone’s hardware, OS, app store and apps, Sovereign GaiXer has four layers: PGX hardware, the Sovereign GaiXer AI-OS, the GaiXer App Store and business apps. The PGX AI server includes Ubuntu, a computer OS, but not an operating foundation for business AI—the AI-OS. With its AI-OS and store, Sovereign GaiXer becomes a generative AI server ready from day one, allowing business apps to be built through its API gateway.
Inside Sovereign GaiXer: the same four-layer idea as a smartphone
A smartphone has four layers: hardware, OS, app store and apps. Sovereign GaiXer follows the same structure.
SmartphoneThe device you use every day
Sovereign GaiXer Your company’s generative AI device
A smartphone’s experience comes from hardware performance multiplied by its OS and apps. Similarly, Sovereign GaiXer’s value combines PGX performance with the AI-OS, app store and apps. The AI-OS turns PGX performance—described as equivalent to 1,000 PCs—into a tool for everyday work.
PGX (hardware)
Lenovo’s compact AI workstation, ThinkStation PGX, is high-performance hardware for running LLMs locally, with performance described as equivalent to 1,000 PCs. The AI-OS helps bring that performance into everyday work.
Sovereign GaiXer(AI-OS)
The AI-OS is an operating foundation for secure business use of AI. As a smartphone OS controls chips and memory, it manages inference engines (such as vLLM and Ollama) and provides defenses against hallucinations and prompt injection. Like a phone’s calling and camera functions, chat comes as a standard app, ready for employees from the first day.
GaiXer App Store
A store for distributing, choosing and using the business apps you build internally. Once an administrator installs apps from the store, employees can use them much as they use smartphone apps.
Business apps
Adding generative AI turns an ordinary business app into a workflow-specific generative AI app. You can use included apps for meeting minutes and presentations, and develop and add your own.
“Can I do the same thing by buying PGX alone?”
The answer is no. More precisely, PGX runs because Ubuntu is installed, but it does not include the AI-OS that makes AI usable for business. The following implementation and operational tasks present a very high barrier, especially for companies without a specialist IT team.
- Managing inference engines (such as vLLM and Ollama): running an LLM requires command-based model deployment, startup and updates, memory management, concurrent-user controls and failure recovery. These day-to-day operations require dedicated engineering resources.
- Hallucination control: measures against plausible but incorrect answers. This is not simply a feature you buy and turn on in settings; the mechanisms for reducing errors must be engineered.
- Prompt injection protection: defenses against crafted instructions that hijack AI and cause information leaks. As attacks evolve daily, ongoing specialist knowledge is needed, rather than a one-time implementation.
- Supporting capabilities for business use: user management, usage logs, app creation and distribution, and other tools for company-wide use must also be built in-house. The next section looks at what they involve.
| Capability | PGX alone | Sovereign GaiXer |
|---|---|---|
| Chat (a standard OS app) | Build in-house | Included |
| App creation and distribution (app store) | Build in-house | GaiXer App Store |
| Management and operation of inference engines (such as vLLM and Ollama) | Build and operate in-house | Managed by the AI-OS |
| Hallucination control | Implement in-house | Managed by the AI-OS |
| Prompt injection protection | Implement in-house | Managed by the AI-OS |
| Users, usage visibility and audit logs | Build in-house | Included |
| Software updates (integrity checks and rollback) | Build in-house | One-click updates |
| Business apps for minutes, presentations and more | Not available out of the box | Available from the store |
PGX alone
- Chat (a standard OS app)
- Build in-house
- App creation and distribution (app store)
- Build in-house
- Management and operation of inference engines (such as vLLM and Ollama)
- Build and operate in-house
- Hallucination control
- Implement in-house
- Prompt injection protection
- Implement in-house
- Users, usage visibility and audit logs
- Build in-house
- Software updates (integrity checks and rollback)
- Build in-house
- Business apps for minutes, presentations and more
- Not available out of the box
Sovereign GaiXer
- Chat (a standard OS app)
- Included
- App creation and distribution (app store)
- GaiXer App Store
- Management and operation of inference engines (such as vLLM and Ollama)
- Managed by the AI-OS
- Hallucination control
- Managed by the AI-OS
- Prompt injection protection
- Managed by the AI-OS
- Users, usage visibility and audit logs
- Included
- Software updates (integrity checks and rollback)
- One-click updates
- Business apps for minutes, presentations and more
- Available from the store
PGX alone versus Sovereign GaiXer (as of August 2026)
Buying only PGX is like getting a smartphone without a phone, camera or app store. It powers on and the screen lights up, but the tools you need every day are missing. The OS and app store turn powerful hardware into an everyday tool. Sovereign GaiXer, the AI-OS, serves that role for generative AI.
The AI-OS does much more behind the scenes
A smartphone OS handles updates, security and usage management out of sight. Sovereign GaiXer does the same. The following are representative capabilities, each substantial enough to require a dedicated development team. Buying an AI-OS means these capabilities are provided from the start.
User and login management
Create and suspend accounts in the admin console, including bulk CSV registration. It separates administrator and ordinary user permissions and supports automatic account locking after repeated failed logins.
Usage visibility and audit logs
Graphs and CSV exports show who uses AI, in which department, and how much. Administrative actions and logins are recorded in audit logs, helping meet IT and audit requirements from the outset.
AI value conversion
Internally processed AI usage is displayed in Japanese yen as what the same work would have cost using cloud AI. Exchange rates are fetched automatically each day, making the value of unlimited use visible to management.
Internal document ingestion (learning sets)
Import PDF, Word, Excel, PowerPoint, images and audio so AI can answer using your internal materials. Tables in PDFs are read as tables.
Speech transcription engine
Transcribe meeting audio entirely on the device without an internet connection. It supports speaker diarization to distinguish who spoke. This is the engine behind the meeting-minutes app.
Software update system
Update with one click in the admin console. Digital signatures verify package integrity, and failed updates automatically roll back to the previous version. Isolated environments without internet access can update using a USB drive.
Automatic clustering from the second unit
In multi-unit deployments, placing a second unit on the same network lets them discover each other and form a cluster automatically. An included monitoring dashboard shows each unit’s health. Initial setup also progresses simply by switching on the power.
With the AI-OS and store in place, you can build apps without developing your own AI foundation
Smartphone users use apps without understanding the OS. Developers build and publish apps without creating an OS of their own. The same principle applies on Sovereign GaiXer.
Build AI-enabled business software through the API gateway
For business software vendors, building and maintaining the AI-OS is a major barrier to an AI-enabled product. On Sovereign GaiXer, their software can call the built-in AI through the API gateway, so they can explore AI-enabled versions running in customers’ on-premises environments.
Early apps are already running
SOAP Assist and Medical Summary Generation are early examples of apps built on this store. Experience in healthcare—a field demanding particularly high security—provides a basis for exploring similar possibilities in other industries and workflows.
Think of it as a company’s generative AI device with its AI-OS (Sovereign GaiXer) and app store (GaiXer App Store) already in place, onto which you add your workflows and software as apps. We welcome conversations about whether your own work or software could become an AI app.
FIXER’s generative AI service is registered as an LGWAN-ASP service
FIXER has operated a generative AI service used by local governments and public institutions.
Registered as an LGWAN-ASP service
FIXER’s generative AI service GaiXer (cloud version) is registered as an ASP service for LGWAN, Japan’s administrative network for local governments (December 2024). Sovereign GaiXer itself is not registered. FIXER has experience supporting secure generative AI use by local authorities and public institutions.
Sovereign GaiXer: frequently asked questions
What is Sovereign GaiXer?
It is an AI-OS, a software foundation for secure business use of AI, that turns an AI server into a generative AI server ready from day one. Like a smartphone made of hardware, an OS, an app store and apps, Sovereign GaiXer consists of PGX hardware, the Sovereign GaiXer AI-OS, the GaiXer App Store and business apps. AI processing stays inside the unit, and you can use chat as well as meeting-minutes and presentation apps.
Can buying PGX alone do the same thing as Sovereign GaiXer?
No. PGX runs because it includes Ubuntu, a computer operating system, but it does not include the AI-OS that makes AI usable for business. Think of a smartphone with no phone, camera or app store. Business use of an LLM requires managing and operating inference engines (such as vLLM and Ollama), controlling hallucinations, and implementing prompt injection defenses. Building and maintaining these capabilities is extremely difficult without a specialist IT team. Sovereign GaiXer includes them as its AI-OS.
What does managing inference engines (such as vLLM and Ollama) involve?
Inference engines (such as vLLM and Ollama) run large language models (LLMs) on a local machine. Business use requires ongoing operation: deploying, starting and updating models through commands, managing memory, controlling concurrent access by multiple users, and recovering from failures. Maintaining this without dedicated engineers is difficult.
What is hallucination control?
A hallucination is an AI response that sounds plausible but is factually incorrect. Business deployments need mechanisms in the inference layer to reduce these errors. There is no ready-made settings screen for building such a system yourself; doing so requires expertise. Sovereign GaiXer implements these controls in its OS-like foundation.
What is prompt injection protection?
Prompt injection is an attack that uses crafted instructions to hijack AI behavior, causing unintended output or information disclosure. Attack techniques keep evolving, so defenses must be updated continuously. Sovereign GaiXer handles these defenses at the platform level.
What does the AI-OS do beyond chat and apps?
Like a smartphone OS handling updates and security behind the scenes, it provides user and login management, usage visibility and audit logs, AI value conversion (the equivalent cloud AI cost in Japanese yen), internal document ingestion (learning sets), speech transcription, digitally signed software updates (including USB updates in isolated networks), and automatic clustering of multiple units.
Who is FIXER, the provider?
FIXER is a Japanese cloud and generative AI company. Its generative AI service GaiXer (cloud version) is registered as an LGWAN-ASP service for Japanese local governments (Sovereign GaiXer itself is not registered).
Can healthcare organizations use it?
Healthcare generative AI apps, including SOAP Assist and Medical Summary Generation, are available through the GaiXer App Store. AI processing stays inside the unit. Whether an app communicates externally differs from app to app. Their output supports drafting and must be reviewed by healthcare professionals.
Can a business software vendor use it to add AI to its own products?
Yes—please contact us. Just as smartphone developers can publish apps without building an OS, Sovereign GaiXer handles inference engines and security. Business software vendors can call the built-in AI through the API gateway, focus on their own application and explore AI-enabled versions running in their customers’ on-premises environments. SOAP Assist and Medical Summary Generation are early examples.
Three points to understand Sovereign GaiXer
- Sovereign GaiXer is an AI-OS analogous to a smartphone’s OS plus app store. Its four layers are PGX hardware, Sovereign GaiXer AI-OS, GaiXer App Store and business apps, following the same structure as a smartphone.
- PGX includes Ubuntu, a computer OS, but not the AI-OS needed for business AI. Implementing and continuously operating management of inference engines (such as vLLM and Ollama), hallucination controls and prompt injection defenses requires a specialist team.
- With Sovereign GaiXer’s AI-OS and store, business apps can be built through the API gateway, and AI-enabled versions of business software become realistic.
Information is based on the Japanese source as of August 2026. The LGWAN-ASP registration applies to FIXER’s generative AI service GaiXer (cloud version), not to Sovereign GaiXer itself. ThinkStation is a Lenovo trademark. “AI-OS” describes the role of the software foundation that supports secure business use of AI. The smartphone OS and app-store comparison is an analogy explaining the product structure, not a claim of compatibility or equivalence with specific OS or store products. “Equivalent to 1,000 PCs” is an indicative description of PGX computing performance, not a guarantee across all uses. Features reflect development status as of August 2026; availability and scope may vary by configuration. Multi-unit clustering features require a multi-node configuration. Healthcare app output is drafting support requiring review by healthcare professionals and does not replace diagnostic or treatment decisions. For overseas availability and applicable commercial terms, contact sales.
Company, product, and service names mentioned are trademarks or registered trademarks of their respective owners.