Redefining desktop AI computing, the Altos BrainSphere™ GB10 F1 is powered by the NVIDIA GB10 Grace Blackwell Superchip from NVIDIA DGX™ Spark, delivering up to 1 PetaFLOP of AI performance in an ultra-compact form factor. Designed for AI model fine-tuning, generative AI, and AI agent development, it brings data center–class computing power to the desktop—enabling fast, efficient, and cost-effective AI innovation without relying on the cloud.
Paired with Altos aiGeni, Altos’ self-developed one-click AI development platform, the GB10 F1 becomes more than a workstation—it’s a complete AI innovation solution. From seamless environment deployment to intelligent resource management and enhanced security, Altos empowers developers, researchers, and enterprises to turn ideas into breakthroughs—faster, smarter, and more securely.
Paired with the Altos aiGeni one-click development platform, users can quickly launch AI projects and seamlessly complete the entire process—from development to deployment—through intuitive, intelligent tools. This integration of hardware performance × smart software delivers a truly all-in-one AI development experience.
| GPU | NVIDIA Grace Blackwell |
| CPU | 20 core Arm, 10 Cortex-X925 + 10 Cortex-A725 Arm |
| CUDA® Cores | 6144 |
| Tensor Core | 5th Generation |
| RT Core | 4th Generation |
| Tensor Performance | 1 petaFLOP AI performance* |
| System Memory | 128GB LPDDR5x, unified system memory |
| Memory Interface | Memory Bandwidth | 256-bit | 273 GB/s |
| Storage | 4 TB NVME.M2 with self-encryption |
| USB | 4x USB4 Type C (up to 40GB/s) |
| Ethernet | 1x RJ-45 connector | 10 GbE |
| NIC | NVIDIA ConnectX®-7 NIC(200G × 2 QSFP) |
| Wi-Fi | Wi-Fi 7 |
| Bluetooth | Bluetooth 5.4 with LE |
| Audio Output | HDMI multichannel audio output |
| Display Connector | 1× HDMI 2.1a |
| NVENC | NVDEC | 1× | 1× |
| OS | NVIDIA DGX™ OS |
| Max Power Consumption | 170W |
| System Dimensions | 150 × 150 × 50 mm(1.13L) |
| System Weight | < 1.5 kg |
| AI Software | Altos aiGeni |
- *Theoretical FP4 PFLOPS using the sparsity feature.