The world’s first dedicated semiconductor product family designed for Physical AI.


Outstanding Real-Time Performance: Instant data processing with no latency
Ultra-Low-Power Design: Extreme efficiency with minimized power consumption
Easy Deployment: Ready for on-site deployment without complex modifications
| Feature | Gen 1 MLSoC | Gen 2 Modalix |
|---|---|---|
| TOPS | 50, optimized for CNN models | 50, with added hardware blocks for GenAI acceleration |
| MLA (ML Accelerator) | Supports INT8 and INT16, Arm A65 × 4 (1.15 GHz) | Supports mixed precision: BFLOAT16, INT8, INT16, Arm A65 × 8 (1.4 GHz) |
| Application Processor | 15K DMIPS | 32K DMIPS, 4 independent CSI-2 interfaces |
| MIPI Support | Not supported (N/A) | 4 × 2.5 Gbps lanes each, Arm C-71 1.2 GHz |
| ISP (Image Processing) | Not supported (N/A) | Supports 8–24-bit RAW, including RGBIR |
| Vision Processor | Synopsys ARC EV-74 800 MHz | Synopsys ARC EV-74 1GHz |
| Video Decoder | H.264/265 4KP60 x 1 | H.264/265/AV1 4KP60 x 1 |
| Video Encoder | H.264 4KP30 x 1 | H.264/265 4KP30 x 1 |
| PCIe | Gen 4 × 8 lanes, Endpoint only | Gen 5 x 8 lanes (Root Complex/Endpoint) |
| Ethernet | 1GbE x 4 | 10GbE x 4 |
| On-Chip Memory | 4 MB | 8 MB, 128-bit, 8 channels, 32/64-bit |
| DRAM | 32-bit LPDDR4 x 4 | LPDDR4 / 4X / 5 |
| Manufacturing Process | TSMC 16nm | TSMC 6nm |
| Supported Models | CNNs, Transformers | CNNs, Transformers, LLMs, LMMs, GenAI |
Hardware development tools designed to help make products such as robots and devices smarter by utilizing SiMa.ai chips.


Maximizing the Potential of Physical AI : Seamlessly integrate Physical AI technology into products through SiMa.ai’s powerful hardware portfolio
Production-Ready : Beyond simple testing tools, SiMa.ai provides high-quality boards that are ready to be applied directly to production products.
Smart and Fast Build : Optimized to help developers reduce trial and error and quickly turn ideas into functional hardware
| Type | Features & Applications |
|---|---|
| SoM (System-on-Module) | Compact board: Used for direct installation in space-constrained environments, such as inside robots and drones |
| PCIe HHHL Card | Card-type board: Used to instantly expand performance by inserting it into a standard PC or server slot |
| Dual M.2 Board | M.2 slot-type board: Used to add low-power AI functionality to industrial mini PCs and similar devices |