The Ryzen AI Max+ Mini PCs and Laptops Heading to IFA 2026

IFA 2026’s pre-show announcements point to a new class of local AI computer: compact mini PCs and laptops with enough shared memory to load substantially larger quantized models than conventional consumer systems. The most notable AMD-based examples include ACEMAGIC’s F9A Pro, Acer’s Aspire G 3D 16, BOSGAME’s M5 MAX, and planned systems from GMKtec, Framework, and Minisforum.

The defining specification is increasingly unified-memory capacity rather than an NPU’s TOPS figure alone. Current announcements commonly target 128GB, while 192GB configurations are beginning to appear. That does not guarantee fast inference or support for every model, but it gives integrated graphics a much larger pool to share with the operating system, applications, and AI workloads.

The headline Ryzen AI Max+ systems

ACEMAGIC F9A Pro: the 192GB mini-workstation preview

ACEMAGIC’s F9A Pro is the higher-end follow-up to the company’s two-litre F9A. It is being teased with AMD’s Ryzen AI Max+ PRO 495, a 16-core, 32-thread Zen 5 processor with boost speeds up to 5.2GHz. Reports describe Radeon 8065S integrated graphics with 40 RDNA 3.5 compute units clocked at up to 3.0GHz, alongside an XDNA 2 NPU rated at up to 55 TOPS and total platform AI performance of up to 131 TOPS. 2324

The major upgrade is memory. The Pro model is reported to support up to 192GB of LPDDR5X-8533 unified memory, with as much as 160GB potentially assignable to graphics. Unlike a conventional laptop with separate system memory and GPU VRAM, that shared pool can be allocated dynamically among CPU, graphics, and local AI workloads. 2324

ACEMAGIC has also been reported to include dual USB4 ports, dual 2.5GbE, Wi-Fi 7, HDMI 2.1, DisplayPort 2.1, and OCuLink over PCIe 4.0 x4. Price, final port configuration, and release timing were not confirmed in the available preview material. 2324

Acer Aspire G 3D 16: a Ryzen AI Max+ laptop with glasses-free 3D

Acer’s Aspire G 3D 16, model AG3D16-81M, uses the Ryzen AI Max+ 395 with 16 cores, 32 threads, boost speeds up to 5.1GHz, and integrated Radeon 8060S graphics. The system supports up to 128GB of LPDDR5X-8000 unified memory and up to a 2TB PCIe Gen 4 M.2 SSD. Acer rates the platform at up to 126 TOPS of overall AI performance. 5158

Its standout feature is a 16-inch SpatialLabs display that can show stereoscopic 3D without dedicated glasses. The IPS panel runs at 3,840 × 2,400 and 120Hz in 2D mode, with 450 nits of brightness and 100% DCI-P3 coverage. In 3D mode, the effective resolution is 1,920 × 2,400. Dual eye-tracking cameras help convert standard 2D material into a 3D presentation without requiring native 3D content or a discrete GPU. 495158

Acer lists a 99.9Wh battery, Wi-Fi 7, Bluetooth 5.4, and two USB-C 4.0 ports. However, the available announcements do not establish the laptop’s weight, battery-life result, regional price, or retail availability. 5154

BOSGAME M5 MAX: a 128GB PRO 495 configuration to start

BOSGAME’s M5 MAX is a preview rather than a fully specified retail launch. The planned first configuration uses AMD’s Ryzen AI Max+ PRO 495 with 16 CPU cores, 32 threads, boost speeds up to 5.2GHz, Radeon 8065S graphics, and an NPU rated at up to 55 TOPS. The announcement gives total platform AI performance as up to 131 TOPS. 38

The initial configuration is planned with 128GB of LPDDR5X unified memory and a 2TB PCIe 4.0 SSD. BOSGAME is positioning the system for local AI experimentation, large-model workflows, creative applications, software development, and professional multitasking. Final ports, dimensions, weight, power limits, price, release date, and additional memory or storage variants were not yet substantiated in the available information. 38

The existing BOSGAME M5 provides a useful baseline

The current BOSGAME M5 shows what the less extreme Ryzen AI Max+ 395 configuration looks like. It combines 16 cores and 32 threads running at up to 5.1GHz with Radeon 8060S integrated graphics containing 40 RDNA 3.5 compute units. Available configurations include 96GB or 128GB of soldered LPDDR5X-8000 shared memory, with up to 2TB of PCIe 4.0 storage. 3940

The M5 includes two USB4 ports, USB-A connectivity, HDMI, DisplayPort, Wi-Fi 7, and an SD card reader. BOSGAME’s own listing shows a 128GB/2TB configuration at $2,999, although prices can vary by retailer and region. 3940

ACEMAGIC F9A: the 395-based model already announced

The standard ACEMAGIC F9A is built around the Ryzen AI Max+ 395, with 16 Zen 5 cores, 32 threads, and boost speeds up to 5.1GHz. Its Radeon 8060S integrated GPU has 40 RDNA 3.5 compute units, while the XDNA 2 NPU provides up to 50 TOPS. ACEMAGIC cites up to 126 TOPS for total system AI performance. 1727

The approximately two-litre system supports up to 128GB of soldered LPDDR5X-8000 unified memory, dual PCIe 4.0 M.2 SSD slots, and an OCuLink port for an external GPU. Its listed connectivity includes two 40Gbps USB4 ports, HDMI 2.1, dual 2.5GbE, Wi-Fi 7, and Bluetooth 5.4. ACEMAGIC markets the 128GB configuration for local inference of LLMs up to roughly 120 billion parameters. 1727

Pricing and a broad retail ship date for the F9A were not confirmed in the cited launch material. A separate report says pre-sales for the 395 configuration were planned for September 10, but regional availability should be verified before treating that date as universal. 23

What unified memory changes for local AI

A large unified-memory pool can be more important than headline TOPS when the goal is to run a model locally. Integrated graphics can draw from the same high-bandwidth memory used by the CPU, allowing systems with 128GB or 192GB to accommodate larger quantized models than many thin laptops with 16GB or 32GB of RAM, or discrete GPUs limited by their VRAM capacity.

The practical caveat is that capacity is not the same as performance. Whether a model runs well depends on quantization, context length, inference software, memory reserved for Windows and other applications, cooling, and sustained power limits. A claim that a computer can load a model does not establish that it will generate responses quickly, run at full precision, or support every model variant.

The evidence is strongest for ACEMAGIC’s own 120-billion-parameter positioning for the F9A and for GEEKOM’s specific demonstration involving four connected A9 Mega systems running DeepSeek V4 Flash locally. The available announcements do not provide comparable measured token-per-second results for the Acer Aspire G 3D 16, ACEMAGIC F9A Pro, or BOSGAME M5 MAX. 5627

Other IFA 2026 systems using the same memory-first idea

GMKtec EVO-X5 Pro

GMKtec has announced the EVO-X5 Pro as a Ryzen AI Max+ PRO 495 mini PC and plans a full reveal at IFA Berlin. The company’s event information confirms the processor focus, but detailed pricing, I/O, cooling, power draw, and final configuration were not yet public. 710

Minisforum T5 Max

Minisforum’s IFA page lists the forthcoming T5 Max with the Ryzen AI Max+ 395, LPDDR5X memory, and two PCIe 4.0 M.2 slots. The page does not provide enough detail to confirm final pricing, memory capacity, thermals, or local-model performance. 16

Framework’s 192GB compact desktop

Framework is applying the same high-capacity approach to a compact desktop configuration based on the Ryzen AI Max+ PRO 495. The reported version offers 192GB of memory and quoted bandwidth of 273GB/s; its Radeon 8065S graphics retain 40 RDNA 3.5 compute units and can reach 3.0GHz. 13

GEEKOM’s distributed local-AI demonstration

GEEKOM is exploring a different way to scale local workloads: connecting four A9 Mega mini PCs through USB4. Each system uses a Ryzen AI Max+ 395 and Radeon 8060S graphics, and GEEKOM says the combined platform can run DeepSeek V4 Flash locally. This is a demonstration of distributed local AI rather than a single compact computer holding the entire workload. 56

How AMD’s systems compare with NVIDIA RTX Spark designs

AMD’s Ryzen AI Max+ systems are not the only IFA products aimed at local AI. Acer has shown an SFF RTX Spark concept with up to 128GB of unified memory and up to one petaflop of compute, while ASUS is presenting RTX Spark-based ProArt P16 and P14 laptops and the GR1X mini PC. ASUS says its platform is designed for local large-language-model, creative, and personal-agent workloads, including models up to 120 billion parameters. 238

The approaches differ. Ryzen AI Max+ combines CPU cores, an integrated GPU, and an NPU in one AMD package, with the appeal of a compact design and a large shared-memory pool. RTX Spark systems emphasize NVIDIA’s CUDA and RTX software ecosystem alongside a dedicated Blackwell-based AI accelerator. Real-world comparisons will depend on software support, memory allocation, sustained power, thermals, and measured application performance—not simply the highest published TOPS or petaflop figure.

What to watch when these systems reach the market

The IFA 2026 announcements show that local AI PCs are moving beyond small NPUs designed mainly for lightweight on-device features. Vendors are now competing on the memory required to load larger models locally: 128GB is becoming common in premium mini workstations, while 192GB is emerging in higher-end configurations.

That does not mean cloud AI is disappearing. Instead, these computers are best understood as another option for privacy-sensitive work, offline use, lower-latency interactions, software development, creative experimentation, and workloads that users prefer to keep on their own hardware. The decisive questions will be whether the promised configurations ship at sensible prices, how well they perform under sustained loads, and which local models and software tools are actually supported.