SiMa.ai ads

Building chips for Physical AI isn’t just about accelerating neural networks—it’s about engineering systems that can interpret, react, and adapt to the real world. That's why we partnerned with TSMC to build Modalix™ MLSoC on the TSMC N6 process.

Featuring seamless open-source or custom LLM import, a curated model zoo, and automated quantization/compilation, LLiMa optimizes open-source or SiMa-precompiled models into Modalix-ready binaries. It enables agent-2-agent systems, Model Context Protocol (MCP), and Retrieval-Augmented Generation (RAG) — all fully on-device for Physical AI.

LLiMa is the industry’s first unified framework to run curated large language, vision, and multimodal models on Modalix—all under 10 watts. Integrate and deploy open-source, custom or SiMa-precompiled models seamlessly.

The era of Physical AI has arrived — and SiMa.ai’s next-gen platform is ready to lead the way. Today, we’re unveiling three major product launches to supercharge the scaling of Physical AI: 1. Modalix™ MLSoC — our next-generation Physical AI silicon is now in production and shipping. 2. Modalix SoM + Devkit — available now and pin-compatible with the leading GPU SoM for fast, seamless integration. 3. LLiMa™ — a new software framework to run high-performance LLM and GenAI models directly on Modalix, without the cloud. This is just the beginning of what’s possible.

Announcing the immediate availability of Modalix System-on-Module (SoM) + Devkit — pin-compatible with the leading GPU SoM for fast, seamless integration.

The era of Physical AI has arrived. Modalix™ MLSoC — our next-generation Physical AI silicon is now in production and shipping.