Framework Desktop brings Ryzen AI Max, 192GB RAM, and local AI power to a compact system

2 Sources

Share

Framework unveiled its latest Desktop at AMD's Advancing AI event, featuring the Ryzen AI Max+ Pro 495 chip with 192GB of unified memory. The mini PC is designed specifically for running large AI models locally, including DeepSeek V4 Flash, and can pool up to 384GB of memory across two machines using high-speed networking.

Framework Desktop Targets Local AI Workloads with Massive Memory

At AMD's Advancing AI event, Framework previewed an upgraded version of its Framework Desktop that pushes the boundaries of what compact systems can achieve for local AI applications

1

2

. The new mini PC is built around the AMD Ryzen AI Max+ Pro 495 chip, an enhanced version of the company's existing processor that delivers 16 full Zen 5 cores with boost clocks reaching 5.2 GHz. Paired with the Radeon 8065S integrated graphics featuring 40 RDNA 3.5 compute units running at 3 GHz, the system offers AMD's fastest integrated graphics yet, capable of handling AAA gaming at 1080p without requiring a dedicated GPU

1

.

192GB of RAM Enables Running Large AI Models Locally

The standout feature is the massive 192 GB unified memory configuration using LPDDR5x RAM operating at 8533 MT/s, delivering 273 GB/s bandwidth

2

. This represents a 50% capacity increase over the previous 128GB models and a 6.6% improvement in memory speeds. Framework states the system is optimized specifically for running large AI models locally, with enough capacity to run DeepSeek V4 Flash at Q8 quantization on a single machine while leaving room for extended context length

2

. For developers and researchers working with LLM workloads, this level of memory eliminates the need for cloud-based inference for many use cases, offering privacy and cost benefits for local AI deployments.

Memory Pooling and Expandability Features

Framework demonstrated a proof-of-concept setup where two Desktop units were connected using 50GbE networking cards, enabling RDMA over Ethernet (RoCE) for ultra-low latency communication

2

. This configuration allows memory pooling across both machines, effectively creating a 384GB unified memory pool through tensor parallelism

1

. The new model also features an open-ended PCIe x4 slot, allowing users to install larger expansion cards that extend beyond standard slot dimensions

2

. Framework will offer the system with Windows or as a Linux-ready pre-built with a distribution already installed, catering to developers who prefer open-source environments for AI development

1

.

Pricing Expectations and Market Position

Framework has cautioned that the 192GB configuration will carry a substantial price premium due to rising LPDDR5X costs. The current 128GB model with the AMD Ryzen AI Max+ 395 costs $3,449, up from $1,999 at its February 2025 launch

1

. Industry observers expect the 192GB model to exceed $5,000, potentially positioning it above NVIDIA DGX Spark, which offers 128GB of memory at approximately $5,000

2

. Framework recommends users consider 32GB, 64GB, or 128GB configurations based on budget constraints. While official pricing and release dates remain unannounced, the system represents a significant option for those seeking to deploy local AI without relying on cloud infrastructure, particularly as concerns about data privacy and inference costs continue to drive interest in on-premises AI solutions.

Today's Top Stories

© 2026 TheOutpost.AI All rights reserved