Gigabyte has finally released their AI Top Aero on November 10, 2026. It's a nifty small AI solution for local AI work. With this mini PC, you can leverage your own hardware for AI models and AI apps.
Compact desktop targets local AI workloads with Blackwell GPU
No need for cloud-based solutions. It's great for developers or creators who want some portable, yet powerful resources for their work.
Specifications
- Processor: Nvidia Grace CPU (20-core, Arm-based)
- GPU: Nvidia Blackwell RTX GPU (6,144 CUDA cores)
- Memory: Up to 128 GB LPDDR5X unified memory
- Storage: Up to 4 TB PCIe 5.0 SSD
- AI Performance: Up to 1 petaflop FP4
Inside, we've got a 20-core Nvidia Grace CPU with a Blackwell RTX GPU with 6,144 CUDA cores. So yeah, it's got the power you need for those intense AI jobs in a smaller package. The whole thing is only about a liter volume-wise. So way more compact than a normal desktop with a tower.
It can be equipped with up to 128 GB of LPDDR5X unified memory and 4 TB PCIe 5.0 SSD. Nvidia says it's capable of up to 1 petaflop FP4 AI performance for all this. Also, it's got USB type-C and 10GbE Ethernet and Wi-Fi 7 for any kind of data connectivity.
Nvidia PAIR is supported for distributed inference as well. It works with stuff like LM Studio and Ollama. First impressions of the underlying RTX Spark platform seem to indicate gaming isn't its most impressive aspect yet.
Some games just don't even work. Others crash or have compatibility issues that can probably be worked on later.
Gigabyte hasn't shared a price yet for the AI Top Aero. But there's a similar AI Top Atom which starts at $4,999 USD (~$4,999). We've covered Nvidia RTX Spark N1X Lags Apple previously.
We've been monitoring Gigabyte launches to get an idea of the base-level capability of the silicon. So today the Aero is globally available and you can get something tangible for local AI development. Despite the software limitations.
Source: NotebookCheck
Discussion
0 comments