ZEEKR 8X with Super Eva Launches, Featuring Step 3.5 Flash Model

ME News reports that on April 17 (UTC+8), 2026, Zeekr officially launched its all-new flagship SUV, the Zeekr 8X, into mass production and market delivery. The vehicle’s intelligent agent, Super Eva, powered by multiple cutting-edge models including Step 3.5 Flash, has successfully completed the critical transition from technology release to large-scale manufacturing and is now being delivered to consumers. This marks a milestone for Step 3.5 Flash, following its open-source debut in March this year, which propelled it to the top of OpenRouter’s global weekly usage rankings—now transitioning from open-source community adoption to widespread commercial deployment in end-user devices. The Zeekr 8X, equipped with Super Eva, is also China’s first vehicle to deliver a “Grok + FSD”-style experience at scale. Over the past year, China’s in-vehicle AI sector has seen considerable buzz around large models being integrated into cars. However, the gap between laboratory demos and real-world, consumer-ready mass production has remained a major industry hurdle. The official launch of Super Eva today confirms that a vehicle-wide intelligent agent powered by Step’s foundational models now possesses the capability for scalable deployment in real-world user scenarios and actual driving conditions. Super Eva was jointly developed by Step, Geely, and Qianli Technology. It is the first vehicle-wide intelligent agent to seamlessly integrate smart cabin, intelligent driving assistance, and digital ecosystems, natively fused with underlying control systems such as autonomous driving, chassis, and powertrain. Its technical architecture closely aligns with Tesla’s Grok-style conversational control system, earning it the title of China’s first “Grok + FSD” system, debuting on the Zeekr 8X. Super Eva integrates three core models from Step. The Step 3.5 Flash inference model is specifically designed for agent scenarios, achieving peak single-request inference speeds of up to 350 TPS. It accurately interprets ambiguous commands and executes multi-step task planning in real time—users simply say, “Take me to pick up my kids after school, stop by a McDonald’s along the way, and arrive at the school before 5 p.m.,” and Super Eva autonomously triggers route planning, intermediate stop scheduling, and activation of driving assistance. Step’s speech large model breaks away from the traditional ASR-LLM-TTS three-stage architecture, delivering lower latency and more natural emotional expression—truly “understanding tone and sounding human.” The vision understanding large model empowers Super Eva with perception of both interior and exterior environments—from assessing road conditions to identifying parking spots—enabling more comprehensive and reliable decision-making. It is understood that Super Eva’s capabilities will continue to evolve and integrate with additional ecosystem services. It will no longer merely be a companion inside the car but will become a seamless intelligent hub connecting users to the wider world. It proactively understands contextual needs—booking restaurants, ordering takeout, purchasing plane tickets, reserving event tickets—and handles all these tasks through natural conversation while driving. Industry experts believe 2026 is a pivotal year for general-purpose large models entering end devices. The mass production and deployment of Super Eva validate the stable deployment capability of models like Step 3.5 Flash in high-real-time industrial terminal environments and provide a replicable path for Step’s deeper penetration into broader hardware scenarios. In the future, Super Eva’s capabilities will expand infinitely—not only as an in-car AI companion but also as an intelligent hub connecting the physical world. It will integrate with Geely’s proprietary after-sales and car rental ecosystems (Geely Ecosystem) and fully incorporate lifestyle services (Digital Ecosystem), eliminating the need to switch between multiple apps or endure cumbersome booking processes. While driving, users need only speak naturally; Super Eva will proactively interpret contextual commands and handle everything from restaurant reservations and food delivery to flight bookings and ticket purchases. (Source: Ifnar)