Zebra Zhixing Unveils “Auto Omni,” an On-device Multimodal AI Stack for Cars—Mass Production Slated for 2026

On Sept. 26, 2025, Chinese in-car software firm Zebra Zhixing launched Auto Omni, a full-modal, on-device large-model solution that it says upgrades the cockpit experience from command-based interaction to “always-on” proactive services. The system is co-developed exclusively with Alibaba’s Qwen team and Qualcomm, and has secured design wins at leading domestic automakers ahead of planned mass production in 2026, the company said at Alibaba’s Apsara (Yunqi) Conference.

Zebra says Auto Omni was built over 15,000+ engineering hours and is delivered as a turnkey package. It bundles: deep customization of Qwen Omni, adaptation to Qualcomm 8397-series SoCs, Zebra’s end-to-end model system, and its Yuanshen AI scenario apps—targeting three base capabilities: multimodal perception, spatiotemporal memory, and all-scenario services.

To bolster content and services, Zebra and partners including NetEase Cloud Music, Ximalaya, Fliggy, Tao Piao Piao and Tmall Auto formed what they call the first “AI In-Car Platform Service Alliance,” covering digital entertainment, local life, and vehicle services. The pitch: deliver platform-level agent products to automakers and smoother in-car interactions for users.

The launch dovetailed with Alibaba’s broader AI push. CEO Wu Yongming reiterated a three-year RMB 380 billion (about $53 billion) infrastructure buildout, and Alibaba Cloud CTO Zhou Jingren introduced seven new models including Qwen3-Max and Qwen3-Omni. Zebra said it is the first to integrate Qwen3-Omni into a production automotive stack.

Embodied/“physical” AI is moving into vehicles as suppliers try to run richer agents on the edge for latency, privacy and connectivity resilience. By pairing Qwen-family models with Qualcomm’s in-car platforms and packaging it as a turnkey solution, Zebra aims to shorten automakers’ path from demo to deployment.