
Minisforum has given businesses more control over local AI data, model hosting and inference with two Ryzen AI Max+ Pro 495 systems unveiled at IFA 2026. The AI Agent NAS N5 Max-P495 stores data and runs AI agents on company hardware, while the MS-S1 Max-P495 mini workstation adds capacity for local models and scalable clusters.
What Minisforum launched at IFA 2026
Minisforum introduced the AI Agent NAS N5 Max-P495 and MS-S1 Max-P495 at IFA 2026, which ran from September 4 to 8 in Berlin. Both products use AMD’s Ryzen AI Max+ Pro 495 processor and integrated Radeon 8065S GPU, with up to 131 TOPS of total AI performance.
The N5 is an upgraded version of Minisforum’s flagship NAS, which launched earlier in 2026 with OpenClaw pre-installed. It can hold up to 200TB of local storage and is intended as a central backend for data, models, knowledge and long-running AI workloads.
Business owners can run agents such as OpenClaw or Hermes Agent directly on the NAS. According to Minisforum, keeping data and inference on local hardware provides privacy, reduces latency and helps control inference costs. The company also notes that agentic AI can use up to 1,000 times more tokens than standard AI.
What the MS-S1 Max-P495 adds
The MS-S1 Max-P495 replaces the Ryzen AI Max 395+ in the earlier MS-S1. Its dedicated NPU provides 55 TOPS, contributing to the shared total of up to 131 TOPS. Minisforum says the system supports models from labs including Gemma, OpenAI and Qwen.
Maximum unified memory rises from 128GB to 192GB compared with the MS-S1 MAX powered by the Ryzen AI Max+ 395, while maximum allocatable graphics memory increases from 96GB to 160GB. That shared memory pool gives a single system more room for demanding models as well as 3D rendering, video editing and computational fluid dynamics.
Minisforum’s October 2026 announcement says DeepSeek V4 Flash 284B in Q4 ran on one MS-S1 Max-P495 at approximately 12 tokens per second. The same announcement reports that two 192GB units can run Qwen3.5-397B locally at approximately 16 tokens per second. Minisforum published the specifications and test results through PR Newswire.
How local AI clusters can support business workloads
The MS-S1 Max-P495 can move from one desktop system to a distributed deployment. Minisforum says four units fit in a 2U rack, and its test configuration successfully ran DeepSeek-R1 671B in Q4_0 quantization across four systems. The model was 380GB, according to the manufacturer’s announcement.
For a business evaluating local AI, this creates options beyond sending every prompt and file to an external service. A single workstation can support local development and inference, a dual-system cluster can handle a larger model, and rack deployment can extend the approach to enterprise or research teams.
Make sure that your agents have proper safeguards so that it doesn’t end up accidentally deleting your entire home directory.
What is available and what still lacks pricing
The MS-S1 Max-P495 with 192GB of RAM and a 2TB SSD is available for pre-order. Minisforum set a regular price of US$9,249 and a 20% discount price of US$7,399, with shipping expected to begin in mid-October 2026. The manufacturer’s announcement contains the pre-order details.
The source coverage did not announce a price or release date for the N5 Max-P495. Availability for that NAS therefore remained unspecified at the time of the IFA announcement.
The MS-S1 is also positioned as an alternative to Apple’s Mac mini and Mac Studio. The source article reported tight supply for those small-form-factor systems during the local AI boom. The 8060S on the older Ryzen AI Max 395+ is also quite capable for 1080p gaming, so we wouldn’t be surprised if you can comfortably play on this new mini-PC, too.
What to do now
- Map the data that should stay local. Identify datasets, model files, knowledge bases and agent instructions that need stronger control over storage and transmission.
- Match the system to the workload. Compare the N5’s storage-first role with the MS-S1’s compute-first design. Account for the required model size, memory, storage and number of connected nodes.
- Request pricing for the complete configuration. The published pre-order price applies to the MS-S1 Max-P495 with 192GB of RAM and a 2TB SSD. Confirm any other memory, storage or cluster configuration before making a purchase decision.
- Plan safeguards before running agents. Give each agent only the permissions and directories required for its work, and protect important files from unintended changes.
- Test against real workloads. The reported DeepSeek, Qwen and DeepSeek-R1 results came from Minisforum’s own tests. Run the intended model and application on your data before depending on it for production work.
FAQ
What processor powers the Minisforum N5 Max-P495 and MS-S1 Max-P495?
Both machines use AMD’s Ryzen AI Max+ Pro 495 processor with an integrated Radeon 8065S GPU, delivering up to 131 TOPS of AI performance.
How much memory can the MS-S1 Max-P495 hold?
The MS-S1 Max-P495 supports up to 192GB of unified memory, with up to 160GB allocatable as graphics memory.
How much storage does the AI Agent NAS N5 Max-P495 support?
The N5 Max-P495 supports up to 200TB of local storage and comes with OpenClaw pre-installed for running AI agents locally.
Sources
See our editorial disclaimer for how our articles are produced.
Run a free scan to see your AI Visibility Score, SEO rating, and local citation accuracy.