AMD and TensorStack Launch Local AI Platform Amuse 3.0
Company AMD presented jointly with TensorStack platform Amuse 3.0designed for local generative AI on the base Ryzen AI processors и Radeon RX video cards. The solution allows users generate images and short videos up to 6 seconds long directly on your devices, without connecting to the cloud.
Amuse 3.0 is based on support for over 100 AI modelsIncluding Stable Diffusion 3.5, FLUX and others. Each model was deeply optimized for AMD architecture, which ensured acceleration up to 4,3 times compared to conventional models.
In comparative tests, the Radeon RX 9070 XT showed that when using AMD Optimized Models (based on Olive + AMD optimizer), image rendering time is significantly reduced:
-
SD 1.5 — up to 4,3x faster
-
SDXL 1.0 — up to 3,1x
-
SDXL Turbo — up to 1,5x
-
SD 3.0 Medium — up to 2,6x
-
SD 3.5 Large — up to 3,3x
-
SD 3.5 Large Turbo — up to 2,1x
In addition, Ryzen AI with a 50 TOPS NPU showed Up to 3,3x faster image generation on optimized models.
The Amuse 3.0 platform also includes AI filters for video and support generating images in typographic quality.
