InfiniCloud® AI 3.0 Revamps Its Product Lineup as a Sovereign ／ Private AI Platform, Offering Solutions Ranging from Enterprise Use to the Free Community Edition (CE)
InfiniCloud® AI 3.0 Revamps Its Product Lineup as a Sovereign / Private AI Platform, Offering Solutions Ranging from Enterprise Use to the Free Community Edition (CE)
New Lineup Includes “Enterprise Control” for Centralized Management of External LLMs, “Enterprise Suite” for Keeping Data In-House, and “LLM Fabric” for Building LLM Infrastructure on Your Own GPUs.
InfiniCloud Co. Ltd. announces the launch of InfiniCloud® AI 3.0, the latest version of its InfiniCloud® AI Sovereign / Private AI platform.
InfiniCloud® AI is a private AI platform designed for enterprises, enabling organizations to incorporate internal policies, documents, and operational expertise into AI, transforming it into an intelligent system that reflects their unique decision-making criteria. The platform allows enterprises to centrally manage and control their AI environments while retaining full ownership and operational control over their AI capabilities.
InfiniCloud® AI 3.0 is a chat-based Agentic AI / AI Agent equipped with personalization capabilities (long-term memory), enabling it to learn from user interactions and adapt to each user’s individual context. Unlike conventional RAG-based retrieval systems that simply search for matching text and often fail to deliver meaningful results, version 3.0 significantly enhances Evidence-Driven Thinking Mode (EDTM), an Agentic AI capability that reads and interprets documents, orchestrates multiple AI models, and derives evidence-based answers.
The latest version also introduces a revamped Search Provider feature, which connects search engines and internal enterprise systems to serve as a centralized information aggregation hub. Its extensibility framework has also been overhauled, including Native Plugins and MCP integration. In conjunction with its AI Agent capabilities, the platform supports extensions through skills, workflows, input adapters, output adapters, and more.
Enhanced Personalization (Long-Term Memory) and EDTM: Delivering Answer Quality Beyond Conventional RAG
InfiniCloud® AI 3.0 remembers individual users' roles, positions, and locations, enabling it to understand the intent behind their questions and provide contextually relevant answers.
For example, a user can simply ask, "Can I fly business class on a business trip to Prague? Am I entitled to a daily allowance?" The AI recalls the user's role, position, and location from previous conversations. It then orchestrates multiple AI models to search across relevant requirements and policies, navigating from executive overseas travel regulations to details per diem allowance tables, ultimately delivering an answer supported by verifiable evidence.
Conceptual diagram. While the diagram reproduces the actual execution sequence, results may vary depending on the conditions. The response has been shortened and restructured based on an actual screenshot. Regional classifications and monetary amounts have been omitted as the have not yet been finalized. Source labels are illustrative and do not constitute direct supporting evidence for the response.
Background
As the adoption of generative AI in business continues to expand, demand for private AI infrastructure is growing, particularly due to security concerns. While generative AI becomes more useful as organizations provide it with access to their confidential data, doing so also increases potential security risks. At the same time, enterprises have varying needs, priorities, and operating environments.
These include:
- Companies that want to make AI available to all employees but struggle to determine what types and levels of data can safely be shared with AI systems.
- Organizations with strict confidentiality requirements that seeks Private AI operated under their own control, or Sovereign AI deployed and operated within their own national borders.
- Organizations that have conducted proof-of-concepts (PoC) trials of RAG but have been unable to achieve the expected results.
- Small and medium-sized enterprises (SMEs) and other organizations that want to explore AI capabilities without incurring upfront costs.
To address these diverse challenges, InfiniCloud® AI 3.0 offers three distinct product series, each designed to meet specific organizational needs.
Product Series Overview
- InfiniCloud® AI Enterprise Control: A Private AI solution that centrally manages requests to external LLM infrastructure (LMaaS) while keeping enterprise information consolidated within on-premises servers or Infrastructure as a Service (IaaS) environments. It can be deployed even in environments without GPUs.
- Supported LMaaS providers include OpenAI, OpenRouter, Sakura AI Engine, and various others. However, services that do not retain data submitted through API requests are recommended.
- InfiniCloud® AI Enterprise Suite: Includes all Enterprise Control capabilities, including integration with external LLM infrastructure, while also providing built-in proprietary LMaaS functionality with fine-tuning support. It enables fully air-gapped deployments on on-premises servers equipped with dedicated GPUs and can also be installed in GPU-enabled IaaS environments.
- InfiniCloud® AI Community: A downloadable software edition designed for small businesses and other organizations, providing access to most InfiniCloud® AI features free of charge under the Elastic License v2 (ELv2), with source code included.
| Features / Specifications | Community Edition | Enterprise Control | Enterprise Suite |
| Product Overview | Free Community Edition | Enterprise AI Integration and Management Platform | Integrated Private AI Platform |
| Target Users | Individuals, developers, and small organizations | Enterprises | Enterprises and organizations handling highly confidential information |
| License | Elastic License v2 | Commercial License | Commercial License |
| Commercial / Business Use | ✔ | ✔ | ✔ |
| Chat-Based Web UI / API | ✔ | ✔ | ✔ |
| RAG / Agentic AI | ✔ | ✔ | ✔ |
| Supported Data Formats | Text, Markdown, Word, Excel, PDF, images, audio, etc. | Text, Markdown, Word, Excel, PDF, images, audio, etc. | Text, Markdown, Word, Excel, PDF, images, audio, etc. |
| AI Agent / Tool Calling | ✔ | ✔ | ✔ |
| Coding Agent Support (e.g., Codex) | ✔ | ✔ | ✔ |
| Extensibility via MCP / Native Plugins | ✔ | ✔ | ✔ |
| External LLM / MaaS Integration | ✔ | ✔ | ✔ |
| Integration with Existing Local LLMs | ✔ | ✔ | ✔ |
| Integrated LLM Inference Engine (Fine-Tuning Support) | ✔ | - | ✔ |
| LoRA Fine-Tuning | ✔ | - | ✔ |
| GPU-Based LLM inference Infrastructure | Included | - | Included |
| Enterprise Authentication and Access Control | Basic Features Only | ✔ | ✔ |
| Organization and Role-Based Access Control (RBAC) | Basic Features Only | Advanced Access Control | Advanced Access Control |
| Enterprise Governance and Auditing | - | ✔ | ✔ |
| Private AI Environment Deployment | Fully Integrated Deployment | Integration with External Inference Infrastructure | Fully Integrated Deployment |
| Operation in Air-Gapped Environments | ✔ | Depends on Connected Infrastructure | ✔ |
| Official Updates and Maintenance Support | Community-Based | Commercial Support | Commercial Support |
- The Community Edition can be used for internal business operations and is source-available. However, it cannot be offered publicly as part of an organization's own service.
- Enterprise Suite includes all the features and capabilities of Enterprise Control.
- For service providers wishing to offer AI services to their customers using Enterprise Control or Enterprise Suite, InfiniCloud offers a dedicated Service Provider Program.
Key Features Enhanced in InfiniCloud® AI 3.0
- Personalization (Long-Term Memory): Retains a complete memory of user conversations, allowing the AI to recall previous interactions, understand the intent behind questions, and provide contextually relevant responses.
- Search Provider: Integrates via APIs with various web search engines, academic research databases, public and specialized databases such as Japan's e-Gov legal database and e-Stat, as well as business card management services. The AI autonomously formulates search queries, retrieves relevant information, and reasons over the results. Because this feature requires access to external APIs, it must be disabled in air-gapped environments. In the Enterprise editions, all data transmitted externally is recorded in audit logs.
- Enhanced Native Plugin Framework and MCP Integration: Provides an extensible framework that allows additional capabilities to be implemented as plugins. Functionality can now be extended across various components, including EDTM processing states, Search Providers, and input/output adapters for Skills and Workflows. The platform also introduces network-based Model Context Protocol (MCP) integration, enabling connectivity with internal business intelligence (BI) systems.
- EDTM v2: A major enhancement to Evidence-Driven Thinking Mode, InfiniCloud's evidence-based reasoning framework. EDTM v2 enables AI to interpret and integrate information from personalization, Search Providers, Native Plugins, and MCP-connected capabilities, using the results as supporting evidence when generating responses. It can also identify information requiring further verification and recommend consulting original source documents. In addition to evidence-driven reasoning, the system supports natural, conversational interactions tailored to each user.
- API Capabilities and AI Coding Agent Support: Enables organizations to issue individual API keys to users, allowing various AI agent harnesses and applications to connect through the platform's APIs. By using InfiniCloud® AI as the backend LLM infrastructure for coding agent harnesses such as Codex and Claude Code, organizations can significantly reduce token-related costs. This capability requires GPU infrastructure and is available in the Community Edition and Enterprise Suite.
InfiniCloud® AI LLM Fabric
InfiniCloud® AI LLM Fabric is a solution-oriented product designed to enable enterprises and service providers with existing GPU infrastructure to launch LLM services (LMaaS) in a short period of time.
Building on the technologies developed through InfiniCloud® AI 3.0, together with cloud infrastructure technologies such as InfiniCloud® HV, LLM Fabric integrates GPU allocation across multiple models, operational management, inference request routing, API key-based access control, and per-user input/output token usage tracking.
The solution provides an integrated framework for transforming GPU infrastructure into a platform capable of continuously delivering LLM services.
Through architecture design, deployment, and system integration tailored to each customer's infrastructure and service requirements, InfiniCloud® AI LLM Fabric supports the launch of commercial LLM API services and shared enterprise AI infrastructure.
Learn More About InfiniCloud® AI LLM Fabric
Fabric: LLM Execution Infrastructure
| Feature | Description |
| LLM API Services | Provides access to multiple models through a unified API endpoint, enabling integration with customer applications and internal enterprise systems (powered by vLLM). |
| API Key Management and Access Control | Controls incoming requests based on API key validity, permitted models, usage limits, and other access policies. |
| Model-Based Routing | Routes requests to the appropriate inference server pool based on the model specified in each request. |
| Inference Load Balancing | Selects the appropriate inference server based on server load and cache conditions within each pool. |
| Inference Server Resource Management | Manages the deployment of multiple models, taking into account model weights, KV cache requirements, runtime memory, and other resource considerations. |
| Inference Server Lifecycle Management | Manages inference server startup and shutdown, inclusion or exclusion from request routing, and graceful shutdown while accounting for in-flight requests. |
| Usage Tracking and Aggregation | Records and aggregates input and output token usage for each request, categorized by user, API key, and model. |
| Deployment Design and Implementation | Designs and deploys solutions tailored to service requirements, including GPU configurations, supported models, access controls, and integration with existing systems. |
Availability and How to Obtain
InfiniCloud® AI 3.0 will be available starting in late October 2026. InfiniCloud® AI Enterprise Control and Enterprise Suite will be offered under subscription-based licenses, available both directly from InfiniCloud and through authorized sales partners.};
InfiniCloud® AI Community Edition will be available for download from the company's website once preparations are complete.};
InfiniCloud® AI LLM Fabric is a solution-based offering that requires customization to accommodate each customer's existing GPU infrastructure.};
Note: The information provided in this press release reflects current plans and is subject to change without prior notice. For further details and pricing information, please contact InfiniCloud.};
For more information about InfiniCloud® AI, please visit: https://infinicloud.com/product_icai.html