AMD democratizes artificial intelligence with local support for OpenAI’s GPT-OSS models

The new Ryzen AI Max+ 395 processor enables running the GPT-OSS 120B model directly from a laptop, marking a milestone in consumer AI computing.

OpenAI has taken a definitive step in opening up advanced artificial intelligence with the release of their GPT-OSS 20B and GPT-OSS 120B models, open-weight language models designed to provide complex reasoning capabilities and local execution. Meanwhile, AMD has announced native support for these models from day one within its hardware ecosystem, positioning its Ryzen AI Max+ processors and Radeon RX graphics cards as the first consumer solutions capable of handling this resource-intensive workload.

Open-Source Models for More Accessible AI

The GPT-OSS (Open Source Sibling) models are the open-weight successors to GPT-2, emphasizing transparency, reproducibility, and local use without dependence on cloud infrastructure. The GPT-OSS 20B model, with 20.9 billion parameters (including 3.6 billion active simultaneously), is designed for demanding real-time natural language tasks, making it more accessible hardware-wise. In comparison, GPT-OSS 120B represents a significant leap, with a total of 116.8 billion parameters and 5.1 billion active—offering comparable capabilities to data center models but capable of running on top-tier personal systems.

AMD Ryzen AI Max+ 395: The First Consumer Processor to Run GPT-OSS 120B

AMD has positioned itself at the forefront of this revolution with its new Ryzen AI Max+ 395 processor, integrated into sleek, lightweight devices like the ASUS ROG Flow Z13. With a total memory of 128 GB (including 96 GB dedicated graphics memory), this processor is the only market solution capable of locally running the GPT-OSS 120B model efficiently for real tasks.

According to internal AMD tests conducted in August 2025, the chip reaches sustained speeds of up to 30 tokens per second, making it a viable option for developers, researchers, and advanced users looking to work with large-scale language models without reliance on remote servers.

AMD Ryzen AI Pro 300

Additionally, support for the Model Context Protocol (MCP) allows maintaining extended conversational contexts and complex workloads, previously only feasible with high-capacity cloud infrastructure.

Full Compatibility with Radeon GPUs for GPT-OSS 20B

For those seeking a more accessible desktop solution, the AMD Radeon RX 9070 XT 16 GB graphics card offers excellent performance in running GPT-OSS 20B. This setup ensures fast generation speeds and significantly reduces the Time To First Token (TTFT), enhancing the experience for interactive applications or local AI agents requiring immediate responses.

Other compatible models include Radeon RX 7000, Radeon RX 9000, and Radeon AI PRO R9000, provided they have at least 16 GB of VRAM.

Running GPT-OSS Locally with LM Studio and AMD

AMD provides a step-by-step guide to download, configure, and run these models easily through LM Studio, the local inference environment. The process involves:

  1. Installing AMD Adrenalin Edition 25.8.1 WHQL drivers or newer.
  2. Adjusting Variable Graphics Memory (VGM) in AMD Software for Ryzen AI processors.
  3. Downloading LM Studio and installing the corresponding GPT-OSS model (20B or 120B).
  4. Enabling GPU inference and manually loading parameters.
  5. Starting chat sessions with full local inference.

Why is this milestone important?

Until now, language models with over 100 billion parameters required infrastructure with specialized GPUs (like NVIDIA A100 or H100), access to platforms such as Azure or AWS, and high operational costs. With this new generation of AMD processors, large-scale AI inference is moving to personal desktops, opening doors for more private, secure, and independent local applications.

Furthermore, this capability strengthens AMD’s commitment to a comprehensive AI strategy, encompassing edge computing, end-user devices, hybrid cloud solutions, and local processing—directly competing with other top tech giants.


FAQs

What is OpenAI’s GPT-OSS 120B model?
GPT-OSS 120B is an open-weight language model with 116.8 billion parameters, designed for complex reasoning, local inference, and advanced generative AI applications, capable of native execution on AMD Ryzen AI Max+ 395 devices.

What hardware is needed to run GPT-OSS 20B or 120B?
For GPT-OSS 20B, a Radeon GPU with at least 16 GB of memory (such as RX 9070 XT) suffices. For GPT-OSS 120B, the AMD Ryzen AI Max+ 395 processor paired with 128 GB RAM and at least 61 GB of VRAM is recommended.

What is the Model Context Protocol (MCP)?
MCP is a technique for maintaining a larger amount of conversational context during LLM execution, improving response coherence and depth. Ryzen AI Max+ processors and compatible AMD GPUs support it.

Where can I download GPT-OSS models, and how are they run?
They are available via LM Studio. AMD has published detailed guides to download models, configure your GPU, and operate the models directly from your PC without needing cloud access.

Scroll to Top