Mercedes‑Benz has started pushing over‑the‑air updates that install Liquid AI’s foundation models on vehicles equipped with third‑ and fourth‑generation MBUX, with an initial October 2026 rollout to some 2027 models, the Wall Street Journal reported and AI Weekly summarised on 5 November 2026.
The partnership behind the update was announced on 23 April 2026, when Mercedes‑Benz said it had entered a multi‑year deal with Liquid AI to "scale embedded, on‑device intelligence" for North American MBUX systems and target an initial production deployment in the second half of 2026. Jörg Burzer, Mercedes‑Benz board member and CTO, said the collaboration aims to "advance on‑device speech, language understanding and reasoning," and Liquid AI CEO Ramin Hasani described the models as built to run on hardware already inside the vehicle.
How the on‑device stack works
Mercedes and Liquid AI say the embedded Liquid Foundation Models (LFMs) run speech recognition, language understanding and even multi‑step reasoning locally on the Mercedes‑Benz Operating System (MB.OS) rather than relying on a continuous cloud round‑trip. The carmaker’s announcement lists the eligible third‑ and fourth‑generation MBUX model years and nameplates. The reported October push targeted the 2027 E‑Class and CLE first, with the 2027 S‑Class, GLC Electric and GLE slated for a second wave in the second quarter of 2027, according to AI Weekly’s summary of the WSJ coverage.
Analysts and explainers note Mercedes positions the embedded models as complementary to cloud‑based large language models: the local model handles latency‑sensitive and offline tasks while the cloud remains available for open‑ended queries, live data and heavy reasoning. The shift reflects a wider trend of smaller, efficiency‑focused LFMs moving to the edge to preserve responsiveness and privacy.
What this means for the market
For Central and Eastern Europe suppliers, dealers and regulators the Mercedes‑Liquid AI move is important even though the initial rollouts are in North America. It signals that one major premium OEM is shipping production‑scale, on‑device conversational models rather than pilots — a step that will increase demand for validated in‑car compute, optimized edge software stacks and secure OTA channels. Vehicle makers and tier‑1 suppliers across CEE that supply infotainment, telematics and domain controllers should expect software and compute requirements to tighten as embedded LFMs spread from high‑end segments into broader lineups.
Regulators and fleet buyers in CEE will also watch how data flows are handled: Mercedes and Liquid AI stress local processing and privacy as advantages, but the companies have not published the precise hand‑off rules between on‑device and cloud processing or the telemetry that will be logged and sent off the vehicle. That leaves compliance, certification and data‑protection questions open — matters of special interest for markets operating under strict EU privacy and automotive cybersecurity rules.
Technically, the rollout underscores two practical points for the region’s auto industry: 1) embedded LLM‑class capabilities can be delivered via OTA to existing hardware generations (so long as the compute and memory budget matches the model footprint), and 2) OEMs are likely to pursue hybrid on‑device/cloud designs rather than replace one with the other. Both change how suppliers design software, validate safety and price software upgrades.
Mercedes Benz models on WorldTrader24 →
Sources: Mercedes‑Benz, AI Weekly (WSJ summary), explainx.ai, Liquid AI, EVinfo.net
