Last Updated on October 6, 2026 by Spencer Lee
AMD says its Ryzen AI Max PRO 400 Series processors can run AI models with more than 300 billion parameters locally when using 4-bit quantization. The platform supports up to 192GB of unified memory, with as much as 160GB available to the GPU, giving high-end laptops far more memory for local AI workloads.
Why is AMD pushing larger local AI models?
AMD is positioning its Ryzen AI Max PRO 400 Series as hardware for businesses and developers that want to run increasingly demanding AI workloads directly on PCs.
The processors combine Zen 5 CPU cores, Radeon graphics and an XDNA 2 NPU with a large unified memory pool. AMD says the platform can run models exceeding 300 billion parameters locally without requiring cloud offload.
The company is targeting AI agents, software development, enterprise workloads and other applications where keeping AI processing on the device can improve privacy, reduce cloud dependence and lower latency.
How much memory do these AI laptops have?
The Ryzen AI Max PRO 400 Series supports up to 192GB of unified memory, with up to 160GB dedicated to the GPU.
That is the key specification for large local AI models. AMD says the new platform increases maximum memory by 50% compared with the previous generation, which topped out at 128GB.
Because the CPU and GPU share the memory pool, the system can give AI workloads access to substantially more memory than many laptops with conventional discrete GPUs.
What does 300B actually mean?
AMD’s 300-billion-parameter claim applies to models using 4-bit quantization.
That does not mean every 300B model will run quickly or perform like a cloud-based system. Quantization, context length, software and the specific model all affect performance.
The more important point for laptop buyers is that memory capacity is becoming a major factor in determining which AI models can run locally at all.
Is the NPU responsible for running 300B models?
Not primarily.
The Ryzen AI Max PRO 400 Series includes an NPU rated at up to 55 TOPS, but AMD’s large-model capability relies heavily on its Radeon GPU and large unified memory architecture.
This is an important distinction as AI laptops increasingly advertise NPU TOPS. A high NPU number can help with supported AI features, but large local LLMs can require far more memory and GPU resources than an NPU alone provides.
Which laptops use Ryzen AI Max PRO?
HP and Lenovo are among the manufacturers bringing Ryzen AI Max PRO 400 Series systems to market.
The HP ZBook Ultra G3a 16, for example, uses the Ryzen AI Max+ PRO 495 and can be configured with up to 192GB of unified memory.
AMD says the platform is also being used in mobile workstations and other professional systems.
What does this mean for AI laptops?
AMD’s latest platform shows that the AI laptop market is beginning to divide into different performance categories.
Mainstream AI PCs can use their NPUs for everyday features such as Windows AI functions, while high-end systems with large unified-memory pools are targeting developers and professionals running much larger models locally.
For those users, RAM and GPU-accessible memory can be more important than NPU TOPS when choosing an AI laptop.
What happens next?
More Ryzen AI Max PRO systems should give buyers more options for local AI workloads as OEMs expand their 192GB configurations.
The next important test will be independent benchmarking of 300B models on retail laptops, particularly token-generation speed, power consumption and sustained performance.
Also read: HP ZBook Ultra G3a Packs 192GB for Local AI Models
HP’s ZBook Ultra G3a is one of the first mobile workstations showing what AMD’s large-memory architecture can do in an actual laptop, including support for very large local AI models.
