AMD has launched the new Stable Diffusion 3 Medium AI model for Ryzen AI laptops. It can generate offline 4-megapixel images and delivers high-quality output with less RAM.
Stable Diffusion 3 Medium: AMD has set a new milestone in the field of Artificial Intelligence. The company recently introduced an AI image generation model that is fully capable of running offline and enables users to generate images up to 4-megapixels from text prompts. This innovative technology is called – Stable Diffusion 3 Medium, which is specifically designed for AMD's Ryzen AI laptops.
What is the Stable Diffusion 3 Medium model?
This new AI model is considered a revolutionary step in AI computing by AMD. This model is optimized for XDNA 2 NPU (Neural Processing Units) and uses the BF16 (Block Floating Point 16) format, allowing it to generate high-quality images with very little RAM. Its special feature is that it is an on-device model, meaning that no internet is required to use it. Data processing takes place entirely on the local device, which not only maintains user privacy but also speeds up processing.
On which devices will this AI model run?
AMD's image generation model will only work on Ryzen AI laptops that have at least 24GB of RAM. In addition, users will also need to download the Amuse 3.1 beta version of Tensorstack, which is the medium for running this model.
What is the image quality and performance like?
The company claims that the Stable Diffusion 3 Medium model can create images of 1024x1024 pixels. Later, with the help of the NPU's processing power, these can be upscaled to 2048x2048 print-ready quality. This gives users the ability to create stock-quality images without the internet, which is especially useful for designers and creators.
Revolutionary improvements in Memory and RAM consumption
AMD has clearly stated that it has focused specifically on the memory optimization of this model. Previously, models like Stable Diffusion XL Turbo required 32GB of RAM, while the new model now runs on only 24GB of RAM. Moreover, it only uses 9GB of RAM during active image generation, making it extremely memory-efficient. This has been made possible by the use of the BF16 format, which delivers better performance even with less memory.
Prompt System: Better Control, Better Output
AMD stated that its model strictly follows the prompt, structure, and order. That is, the user has to first give the subject, then the structural elements, and finally the details while generating the image. This gives the model more accurate and desired results. In addition, Negative Prompting has also been supported, allowing you to remove unwanted elements from the image. It is worth noting that the full stop at the end of the prompt can also affect the model's understanding of context.
Tensorstack Amuse 3.1 App: New AI Hub
The new Stable Diffusion model is part of the Amuse 3.1 desktop app developed by Tensorstack, which can be downloaded for free for AMD Ryzen AI laptops. This app is not only the medium for running this model, but it also includes other AI tools that can take creativity to new heights.
Why is this technology special?
- Internet Free Model – Data is processed entirely on the device.
- Better Privacy – No need to upload to the cloud.
- High Quality in Less RAM – Excellent output in less memory due to BF16 format.
- Print-Ready Output – Photos up to 4 megapixels possible.
- Prompt-based control – User gets more control over the output.
