AMD and Qualcomm Officially Announce Support for GPT-OSS Model

AMD and Qualcomm have jointly announced that their latest hardware platforms now support OpenAI’s newly released gpt-oss series of open-source inference models, marking a significant advancement in edge computing and on-device artificial intelligence. The gpt-oss series includes two models: gpt-oss-20b, a more lightweight version, and gpt-oss-120b, a more complex and powerful variant. The gpt-oss-20b model can run smoothly on devices with just 16GB of memory, while gpt-oss-120b is optimized to operate efficiently on a single 80GB GPU. AMD highlighted that the Ryzen AI Max+395 processor is now the world’s first consumer-grade AI PC chip capable of running the gpt-oss-120b model. To enable this performance, AMD leveraged the GGML framework and MXFP4 precision format, allowing the model to run effectively using approximately 61GB of VRAM. The company’s “Strix Halo” platform, equipped with 128GB of unified memory, can allocate up to 96GB to the GPU, meeting the demanding requirements of the large model. In performance benchmarks, the Ryzen AI Max+395 delivers up to 30 tokens per second when running gpt-oss-120b, with support for the MCP model context protocol, enabling faster response times and improved efficiency during complex AI tasks. Qualcomm confirmed that early testing of gpt-oss-20b on its Snapdragon platforms has demonstrated strong reasoning capabilities, particularly in chain-of-thought inference. Developers can now access the model through popular platforms like Hugging Face and Ollama on Snapdragon-powered devices, making it easier to deploy and experiment with the model. This collaboration underscores the strategic vision of both AMD and Qualcomm in advancing AI at the edge. By enabling powerful open models to run directly on consumer hardware, they are paving the way for smarter, more responsive devices and unlocking new possibilities for AI applications across personal computing, mobile, and embedded systems. As adoption of the gpt-oss series grows, users can expect increasingly intelligent and efficient AI experiences in everyday devices.