Conversational AI-powered user experience pioneer Cerence AI and ultra-efficient MLSoC (machine learning system-on-chip) expert SiMa.ai today announced a collaboration that brings CaLLM Edge, Cerence’s automotive-grade embedded SLM (small language model), to SiMa.ai MLSoC Modalix SoM (system-on-module) to deliver more intelligent, low-power in-car conversational AI experiences in vehicles. The joint solution will debut at IAA Mobility 2025 in Munich, where Cerence will showcase Cerence xUI, the company’s agentic AI assistant platform, with CaLLM Edge running on SiMa.ai hardware.

“Our partnership with SiMa.ai represents a major leap forward in deploying efficient on-device AI for current and next-generation vehicles,” said Nils Schanz, Chief Technology & Product Officer at Cerence AI. “Within Cerence xUI, CaLLM Edge delivers advanced reasoning, multi-turn conversation, and proactive engagement, all running seamlessly on the MLSoC Modalix. With cross-platform compatibility and hardware flexibility, we empower OEMs to adapt quickly, reduce complexity, and deploy bespoke in-car assistants across different vehicle platforms.”

The integration of CaLLM Edge with SiMa.ai’s Modalix platform delivers on the needs of automakers and end users by providing a robust, high-performing technical foundation for in-vehicle AI at the edge. Cerence’s hybrid AI creates efficient and fluid transitions between edge and cloud execution, a critical requirement in delivering an uninterrupted, high-quality conversational user experience. By leveraging the edge capabilities of xUI on Modalix, OEMs can optimize the user experience by delivering real‑time, multi-modal interaction and processing sensor inputs as well as conversational inference directly on-device with claimed industry-leading power efficiency.

Today’s drivers expect features to work everywhere, putting pressure on OEMs to deliver smart, reliable experiences. However, cloud-only solutions rely heavily on connectivity and can result in degraded user experiences during connectivity loss.

Cerence’s hybrid AI architecture tackles these challenges by combining embedded AI for ultra-low-latency responsiveness and enhanced privacy with enriched, context-aware information from the cloud. With CaLLM Edge running on Modalix, the in-car AI experience for the end user is faster, more responsive, more conversational, and available regardless of connectivity.

For many automakers, increased flexibility and resilience are key considerations. As OEMs face increasing platform complexity and navigate supply chain volatility, there is a growing need for hardware flexibility and cost-effective performance.

CaLLM Edge is engineered for cross-platform compatibility and built to run efficiently on a range of hardware architectures. With the addition of MLSoC Modalix to Cerence AI’s platform ecosystem, OEMs now have access to a compact, power-efficient, and scalable edge solution that enhances flexibility in deployment.

“The ultimate vision is voice control that understands the user like a human would,” said Harald Kröger, President of Automotive at SiMa.ai. “LLMs offer this, but until now, it was only possible via the cloud or with power-hungry graphics cards that make it practically impossible to use in cars or any kind of machine.”

He says that the SiMa.ai/Cerence AI solution is “a production-ready device with extraordinary performance and energy efficiency.” It is paired with the LLiMa toolchain, which he says dramatically reduces integration effort and development time.

It was just last month that Futurride reported that SiMa.ai’s next-gen physical AI platform reached production. In addition to the San Jose, CA-based startup announcing Modalix for physical AI applications is in production and available for customers, it released a system-on-module and devkit to accelerate production and debuted the LLiMa software framework built to deploy LLMs and GenAI models.

Leveraging decades of innovation and expertise in voice, generative AI, and large language models, Burlington, MA-based Cerence powers integrated experiences that create safer, more connected, and more enjoyable journeys, with more than 525 million cars shipped with its technology.

Last month, the company announced the next evolution of its collaboration with the Volkswagen Group. The companies enhanced the automaker’s IDA in-car assistant to now enable faster, more natural conversations with a more helpful, knowledgeable companion ready to answer questions, suggest ideas, help with navigation, share fun facts, and tell a story to entertain children in the car.

“By upgrading our in-car assistant in partnership with Cerence AI, we are delivering even more human-like conversations and a vastly expanded knowledge base, making every journey more enjoyable and connected—and delivering against our promise of continuously bringing added value and new capabilities to our drivers,” said Kai Grünitz, Member of the Volkswagen Brand Board of Management responsible for Development.

The upgraded IDA experience—already available in most recent models across various Volkswagen Group brands and models—now supports more complex, multi-turn conversations, powered by Cerence Chat Pro. That means drivers can follow up naturally without needing to repeat themselves or rephrase their request, making the experience more conversational.

“This advanced user experience is a clear illustration of Cerence AI and Volkswagen Group’s shared commitment to making the in-car experience more personal, connected, and intelligent—addressing real driver needs and delivering added value even after vehicle purchase,” said Christian Mentz, Chief Revenue Officer for Cerence AI. “Cerence AI and Volkswagen are together defining the future of in-car interaction: conversational, intelligent, and deeply integrated. We are proud to build on our strong history of collaboration with Volkswagen to further support their commitment to innovation and driver-centric technology.”

In May, Cerence AI announced it was teaming with Arm to optimize real-time GenAI performance in cars without relying on GPU-intensive platforms. The collaboration leverages Arm’s Kleidi software to enhance Cerence’s CaLLM Edge, allowing GenAI workloads to run agentically on Arm CPUs, which are already present in 94% of infotainment systems on the road today. This not only boosts performance and privacy but also opens the door to democratized, multimodal in-car AI in lower-cost vehicles, which typically require more compute power than most CPU systems can support.