9 June 2025 | Engineering
Driving the World, Sight Unseen: Wayve’s AI-500 Roadshow Hits 90 Cities in 90 Days
One AI model. 90 cities. Zero retraining. In the first legs of our AI-500 Roadshow, Wayve’s foundation model demonstrated how general-purpose driving intelligence can scale globally, adapting to unfamiliar roads, traffic, and weather conditions across three continents.
At Wayve, we’re building a general-purpose AI Driver—a foundation model trained on large-scale, diverse driving data to unlock global autonomy. Like humans, our model doesn’t memorise routes or rely on HD maps. It understands the world, adapts to new environments, and applies learned experience to novel situations.
To prove this, we launched the AI-500 Roadshow—our most ambitious real-world test yet. The goal: take a single model to 500 cities by the end of 2025. No retraining. No region-specific coding. Just one model, tested on real roads around the world.
We’ve just completed the first three legs: 90 cities in 90 days, spanning Europe, North America, and Asia. Here’s what we’ve learned about building driving intelligence that scales globally.
A new benchmark for generalization
This milestone marks our most extensive demonstration of Embodied AI for driving, showing how a foundation model can generalise to new, diverse, and previously unseen environments.
Of the 90 cities:
- 62% were new to our test fleet: we had never deployed the AI Driver in these places before.
- Instead, our model applied the general concepts of driving to new scenarios, like hairpin bends in the Swiss Alps.
To date, the roadshow has logged over 10,000 hours of supervised AI driving, half of it in dense urban areas. From storm-soaked highways in Montana to tractor trails in rural Italy and the narrow backstreets of Yokohama, our AI performed without HD maps, geofencing, or region-specific logic.
From London to Schwyz, Chiyoda to Varese, the AI-500 Roadshow put our foundation model to the test as it navigated a wide range of cities, road types, and driving conditions.
How Our Foundation Model Drove in 90 Cities Across 3 Continents
What’s remarkable is that we didn’t fine-tune the model for any region. The same foundation model powered every drive, handling unfamiliar road layouts, varied weather, and unique traffic systems with ease.
That adaptability comes from how we train. Our model is built on a rich, multimodal dataset collected by internal fleets in London, the Bay Area, Stuttgart, and Tokyo, combined with partner data from automakers and third-party fleets.
Each partner provides only a small amount of data, which alone isn’t enough to train a capable driving model. But when federated with Wayve’s dataset, those contributions enrich the model’s understanding of global driving.
We saw this data-to-value advantage in action:
- 47% of cities had only sparse third-party or OEM data
- 14% were entirely novel, with zero prior data
This is the strength of our federated learning approach: combining fragmented datasets into a single, intelligent model that can drive anywhere. For automakers, this means faster deployment of AI driving systems that scale globally, with minimal localisation.
Tested Across Diverse Operational Design Domains
The AI-500 Roadshow was designed to push the limits of our model across a broad range of ODDs. Here’s how the first 90 cities break down by continent:
By Time & Weather:
We tested across different lighting and environmental conditions.
By Road Type:
From narrow mountain roads to high-speed highways.
Driving Intelligence Built to Scale
This is what global generalisation looks like: a single foundation model learning from globally distributed data and applying that knowledge anywhere. Every new city reinforces our conviction that one model can be scaled across continents, vehicle platforms, and driving cultures. This is the foundation for automotive-grade AI.
Wayve’s product vision is to empower automakers with AI driving technology that adapts to any vehicle and any environment, enabling assisted and automated driving to scale across global markets.
What’s Next
The roadshow continues. Next, we head to the US East Coast, followed by new routes across Europe and Japan—each presenting fresh challenges and opportunities to advance our model. Stay tuned as we continue pushing the boundaries of Embodied AI and share more from the road.
👉 Watch the latest video from our drive through the Alps, from Milan, Italy to Stuttgart, Germany.
3 August 2026 | Research
GAIA-4: Multimodal World Models Powering Closed-Loop Simulation for Safe and Scalable Autonomy
The latest evolution of GAIA, GAIA-4, puts the Wayve AI Driver in the loop, transforming world modeling from scene generation into a way to measure end-to-end safety at scale.
Read more
1 July 2026 | Life at Wayve
Wayve Announces $85M Employee Tender Offer
Wayve is offering an $85 million employee tender to give current employees liquidity, as we continue executing on our ambitious long-term roadmap.
Read more
29 May 2026 | Research
Building Intelligence That Can Act in the World
Frontier AI has made extraordinary progress by scaling digital intelligence. Much of the field is focused on making these systems more general-purpose, autonomous and useful across the digital world. Wayve Labs is focused on a different frontier: embodied AI.
Read more