AMD has launched Amuse 3.0, a new local generative Artificial Intelligence platform developed in collaboration with TensorStack AI, targeting users of AMD Ryzen AI processors and Radeon RX GPUs. The platform enables the creation of both print-quality images and draft-quality short videos—up to 6 seconds in length—without the need for cloud-based resources or external data transfer, thus ensuring privacy and speed for end users.
Amuse 3.0 offers compatibility with over 100 advanced Artificial Intelligence models, including the latest versions of Stable Diffusion (3.5) and FLUX. These models have been meticulously optimized for AMD hardware, delivering up to a 4.3 times improvement in inferencing speed compared to generic alternatives. AMD achieved these performance gains by benchmarking optimized models against standard Olive Optimize base models running on the AMD Radeon RX 9070 XT GPU. The enhancements also extend to Ryzen AI processors equipped with a 50 trillion operations per second (TOPS)-class neuroprocessing unit, where image generation showed a 3.3x speed boost.
In addition to faster image and video generation, Amuse 3.0 introduces Artificial Intelligence-powered video filters, further broadening creative possibilities for users. The breakthrough local capability and hardware adaptations target artists, content creators, and professionals seeking real-time performance and autonomy. AMD presents Amuse 3.0 as a significant step toward localized, high-efficiency generative Artificial Intelligence on consumer-grade hardware, reflecting ongoing efforts to advance both hardware and model-level innovation in the field.
