Last week, I saw that Zeekr announced the start of mass production of their new flagship SUV, the Zeekr 8X.

This in itself is noteworthy, but what’s even more interesting is the presence of the AI agent called "Super Eva" installed in this vehicle.

In simple terms, Super Eva is an AI system centered around Zeekr’s Step 3.5 Flash model, allowing all in-car operations to be controlled through natural conversation alone.
It’s not just an extension of voice recognition; it understands and executes complex multi-step tasks. For example, by simply saying, “Pick up the kids from school, stop by McDonald’s on the way, and arrive at school by 5 o’clock,” it automatically plans the route and identifies parking lots.
This is a significantly challenging area at the mass production Eva stage.

As background, Zeekr’s Step 3.5 Flash has been open-sourced since March and has ranked at the top globally in weekly requests on OpenRouter.
In other words, the industry is at a major turning point, transitioning from demo experiments in labs to actual commercial deployment for end users.
This mass production Eva deployment proves that the foundational model can truly operate at scale.

Super Eva is a joint development by Zeekr, Geely, and Qianli Technology, integrating smart cabin, smart driving assistance, and digital ecosystem.
It’s called the “First Chinese Grok + FSD experience” because it aligns with Tesla’s Grok conversational architecture.
Step 3.5 Flash achieves inference speeds of up to 350 TPS per request and can accurately understand vague instructions.
The voice large model also moves away from the traditional ASR-LLM-TTS three-stage setup, reducing latency and enabling more natural emotional expression.

Personally, I feel that this transition to the mass production Eva stage symbolizes the overall trend in the AI industry this year.
Industry insiders also believe 2026 will be the key year for the full integration of general large-scale models into end devices, and the deployment of Super Eva has become a concrete example of that.

What’s interesting is the future scalability.
It’s planned to evolve from a mere in-car AI partner to a smart hub connected to the physical world.
If integrated with Geely’s after-sales service and rental car ecosystem, a world will come where tasks like restaurant reservations, delivery orders, and flight bookings can be completed through natural conversation while driving.

As more examples of mass production Eva implementations emerge, the way AI technology is socially integrated will likely undergo significant changes.