SiMa.ai has launched its second-generation MLSoC platform, Modalix, designed to operate large language models (LLMs) and generative AI workloads on edge devices using less than 10 watts of power. The platform includes a pin-compatible SoM, a DevKit, and the LLiMa software framework, now available for immediate enterprise deployment.
Key Features of the Modalix MLSoC Platform
- Ultra-Low Power Consumption
The Modalix platform runs advanced AI models—including LLMs, transformers, CNNs, and vision-language models—on under 10W, eliminating thermal throttling while enabling real-time inference. - Plug-and-Play Integration
The System-on-Module (SoM) form factor maintains pin compatibility with leading GPU SoMs, making integration into existing hardware systems straightforward. - LLiMa: Edge-First AI Software Stack
LLiMa automates the end-to-end deployment workflow—from model import to quantization, compilation, and runtime orchestration—enabling complex GenAI models to run on Physical AI systems without cloud dependency. - Broad Industry Readiness
Modalix is purpose-built for sectors like robotics, industrial automation, automotive, healthcare, retail, and smart vision—effectively enabling AI “reasoning” inside edge devices. - Production-Ready Availability
Starting prices are $349 for the 8 GB SoM, $599 for the 32 GB version, and $1,499 for the full DevKit, available now for enterprise deployments.
Why It Matters
- Edge AI Without Trade-offs
Under-10W power usage makes Modalix ideal for battery-powered or thermally constrained applications, shifting LLM-powered reasoning to devices like robots, cameras, and vehicles. - Simplified, Reliable AI Deployment
The combination of hardware and automated software (LLiMa) transforms what once required specialized engineering into a turnkey development experience. - Secure and Responsive Use Cases
By keeping AI inference local, Modalix ensures low latency and improves data privacy compliance—critical in regulated industries. - Physical AI Becomes Real
With this platform, SiMa.ai is pushing the concept of Physical AI—where machines perceive, reason, and act using on-board LLMs—into real-world production.
