AMD Unveils 'RyzenClaw' & 'RadeonClaw': Empowering Local AI Agents on PCs, Beyond the Cloud

AMD Unveils 'RyzenClaw' & 'RadeonClaw': Empowering Local AI Agents on PCs, Beyond the Cloud

In a strategic move underlining its commitment to shaping the future of computing, AMD has announced the launch of two groundbreaking initiatives: 'RyzenClaw' and 'RadeonClaw'. These represent a significant shift towards enabling the execution of Artificial Intelligence (AI) Agents locally on personal computers, moving away from traditional reliance on cloud infrastructure. This endeavor aligns with AMD's vision to provide comprehensive solutions for developers and researchers, allowing them to explore the full potential of AI at the edge.

This direction highlights AMD's foresight for smarter, more independent personal computers. The company has released a detailed guide for running the OpenClaw platform on Windows, fully leveraging its advanced hardware. These configurations facilitate the local execution of high-specification AI models, ensuring ultra-fast response times and enhanced data privacy, while reducing the need for constant internet connectivity or dependence on cloud data centers.

RyzenClaw: Integrated AI Processing Power

The RyzenClaw pathway is built upon the Ryzen AI Max+ platform, which boasts an impressive 128GB of unified memory. A substantial 96GB of this memory is dedicated to variable graphics memory (VRAM), providing unprecedented capability for running complex AI models. Initial benchmarks demonstrate remarkable performance; this setup can process the Qwen 3.5 35B model at speeds of up to 45 tokens per second, while accommodating a massive context window of 260,000 tokens.

RyzenClaw's capabilities extend beyond mere speed to include efficiency in handling demanding tasks. It can process 10,000 tokens in approximately 20 seconds, and more impressively, it supports running up to six concurrent AI agents. This opens new horizons for multi-agent AI applications and complex tasks requiring coordination among several agents.

RadeonClaw: Superior Graphics Performance for AI

Conversely, the RadeonClaw pathway focuses on superior graphics processing power, leveraging the Radeon AI PRO R9700 graphics card. This card comes equipped with 32GB of dedicated graphics memory (VRAM) and offers even greater token processing speed, reaching up to 120 tokens per second. This translates to the ability to process 10,000 tokens in a remarkably short time of about 4.4 seconds, making it ideal for tasks demanding immediate response and rapid data processing.

While RadeonClaw may be limited to a context of 190,000 tokens and supports only two concurrent agents, its superior speed makes it an excellent choice for developers seeking maximum performance in natural language processing and graphically intensive AI tasks.

Target Audience and Practical Application

Through these solutions, AMD targets a specific segment of users: developers and AI enthusiasts who wish to explore local execution models for AI tasks. This environment offers an alternative to enterprises relying on large data centers, potentially reducing operational costs and enhancing data security and privacy.

The company clarified that the technical environment for these solutions relies on Windows Subsystem for Linux 2 (WSL2) and the LM Studio tool for local model execution, utilizing the popular llama.cpp library. These platforms also provide the capability for local embedding and memory storage, eliminating the need for any external connection and ensuring complete operational independence.

Cost and Future Directions

Given the advanced specifications and professional capabilities, the cost of these devices places them in the category of professional-grade hardware for advanced experimentation. The RyzenClaw platform starts at $2,700 USD, while the Radeon AI PRO R9700 card alone is priced at $1,299 USD. These price points currently confine these systems to research and development circles, at a time when the mainstream consumer market for local AI is still in its nascent stages.

These initiatives arrive amidst growing interest in local AI, coinciding with previous news regarding OpenAI's recruitment of an OpenClaw developer, underscoring the importance of these technologies in fostering the next generation of personal AI agents. As development continues, we may foresee a future where the reach of these solutions expands to include a broader range of users, paving the way for a new era of independent AI-powered personal computing.