Minisforum is expanding its local AI hardware lineup with two new Mini PCs built around AMD’s Ryzen AI MAX+ PRO 495 processor. The MS-S1 MAX-P495 AI Mini Workstation and N5 MAX-P495 AI Agent NAS are designed to run demanding AI workloads locally, with the company targeting applications that require high memory capacity, fast inference, and large amounts of stored data.
The hardware is being presented at IFA 2026 in Berlin, where Minisforum is positioning the two machines as complementary parts of a local AI platform. The company claims up to 131 TOPS of AI compute performance, with support for up to 192GB of memory at 8533 MT/s and as much as 160GB of graphics memory.
The N5 MAX-P495 is the storage-focused system of the pair. It combines NAS functionality with local AI processing and can accommodate up to 200TB of storage. Minisforum sees it as a central location for AI models, datasets, embeddings, databases, and RAG knowledge bases, allowing these resources to remain on a local network rather than being sent to cloud services.
The MS-S1 MAX-P495, meanwhile, is aimed at users who need more compute power in a compact workstation. Minisforum says it is designed for AI model inference, development, computer vision, generative AI, and other workloads where low-latency local processing is important.
The two systems can also be used together. In Minisforum’s proposed setup, the MS-S1 handles compute-heavy applications and real-time inference, while the N5 provides storage and persistent AI infrastructure. This gives users a way to separate AI compute from data storage while keeping both within a local environment.
Minisforum is demonstrating the new systems at IFA 2026 from September 4 to September 8 in Hall 6.2-104. The company has not yet announced pricing or full retail availability for either model.
| Specification | Minisforum MS-S1 MAX-P495 | Minisforum N5 MAX-P495 |
|---|---|---|
| Product Type | AI Mini Workstation | AI Agent NAS |
| Processor | AMD Ryzen AI MAX+ PRO 495 | AMD Ryzen AI MAX+ PRO 495 |
| AI Performance | Up to 131 TOPS | Up to 131 TOPS |
| Memory | Up to 192GB | Up to 192GB |
| Memory Speed | 8533 MT/s | 8533 MT/s |
| Graphics Memory | Up to 160GB | Up to 160GB |
| Local Storage | Not specified | Up to 200TB |
| Primary Focus | High-performance local AI computing | AI storage + computing |
| AI Inference | Real-time local inference | Local AI processing |
| AI Agents | Supported | OpenClaw, Hermes, and other AI agents |
| RAG / Knowledge Bases | Supported | Supported |
| Databases & Embeddings | Supported | Supported |
| Generative AI | Supported | Supported |
| Computer Vision | Supported | Supported |
| Persistent AI Workloads | — | Yes |
| Role in AI Infrastructure | AI compute frontend | AI compute and storage backend |
| Best Suited For | AI development, inference, engineering and edge workloads | AI models, datasets, RAG, databases and long-running AI workloads |
| Connectivity / Expansion | Extensive connectivity and expansion | Designed for centralized local storage and AI infrastructure |
| IFA 2026 Location | Hall 6.2-104 | Hall 6.2-104 |
