SiMa.ai, a leader in Physical AI solutions, today announced the production and immediate availability of its second-generation Machine Learning System-on-Chip (MLSoC™) – Modalix™ – designed to accelerate the scaling of Physical AI across industries.
As sectors such as robotics, autonomous vehicles, industrial automation, and aerospace increasingly push AI to the edge, they face a common challenge: achieving high performance within the strict power, size, and integration constraints of edge devices. Cloud-based AI often falls short due to latency and high energy consumption. Modalix™ addresses this gap by delivering high performance and accuracy under 10 watts, capable of running LLMs, transformers, CNNs, and GenAI workloads efficiently.
Performance, Flexibility, and Low Power
Built on a flexible Arm-based architecture with a native GenAI software stack, Modalix™ supports real- time perception, decision-making, and natural language interaction. Its compatibility with key interfaces such as camera, Ethernet, and PCIe makes it adaptable for use in robotics, automotive, industrial automation, aerospace and defense, smart vision, retail, and healthcare applications.
Complete Platform for Physical AI
Alongside Modalix™, SiMa.ai introduced:
- Pin-Compatible System-on-Module (SoM) – Developed with Enclustra, the compact, power-efficient SoM offers a drop-in replacement for leading GPU SoMs, integrating MIPI, memory, and essential I/O for rapid deployment.
- LLiMa™ Framework – A unified on-device platform for running LLMs, LMMs, and VLMs entirely offline, with features such as curated model zoo access, automated quantization/compilation, and support for agent-to-agent systems, MCP, and RAG.
The integrated Palette™ SDK software, enables developers to move from prototype to production quickly and cost-effectively.
Industry Partnerships Driving Innovation
“SiMa.ai’s Modalix showcases the scale of innovation possible on Arm’s flexible, high performance, power-efficient compute platform,” said Ami Badani, Chief Marketing Officer, Arm. “By bringing AI and LLM capabilities to Physical AI applications at the edge, SiMa.ai is enabling smarter, faster, and more sustainable systems across industries.”
“The development of Physical AI applications requires validated, purpose-built silicon and software, only possible using advanced design solutions,” said Ravi Subramanian, Chief Product Management Officer, Synopsys. “Achieving a successful first tapeout of MLSoC Modalix illustrates the mission-critical role of Synopsys AI-powered design and IP.”
“This Enclustra–SiMa.ai SoM is more than just a module – it’s a ready-to-deploy Physical AI platform,” added Philipp Baechtold, CEO of Enclustra.
Leadership Perspective
“The era of Physical AI is here,” said Krishna Rangasayee, Founder and CEO of SiMa.ai. “With Modalix™ now in production, we’re accelerating its global adoption and simplifying on-device LLM deployment. Demand for our Modalix SoM is strong, and we’re enabling developers worldwide to bring GenAI to Physical AI systems faster than ever.”
TSMC’s advanced N6 process technology powers Modalix™, ensuring it meets stringent embedded power, thermal, and reliability demands. “TSMC is proud to collaborate with SiMa.ai to deliver advanced SoCs that meet the growing demand for Physical AI,” said Sajiv Dalal, President of TSMC North America.