Quick Takeaways
- Rivian unveiled a major in-house AI compute and multi-sensor autonomy stack aimed at scaling hands-free driving toward higher SAE levels.
- Gen 3 autonomy hardware with lidar and RAP1 compute is targeted to debut with the R2 SUV by end-2026, alongside a paid Autonomy+ service.
On December 11, Rivian revealed major updates to its Rivian autonomous driving platform, outlining how new AI-driven software and next-generation vehicle hardware will reshape hands-free driving across its future electric vehicles. The announcements were made during the company’s Autonomy & AI Day in Palo Alto, highlighting Rivian’s long-term vision for scalable, safe, and data-driven autonomy.
Rivian confirmed that its Gen 2 vehicles and upcoming Gen 3 models will feature a significantly upgraded autonomous sensor and compute architecture. For the first time, the system combines high-resolution cameras, multiple radar units, and lidar, creating a multi-modal sensing approach designed to enhance real-time perception, redundancy, and operational safety.
Rivian autonomous driving platform built on in-house AI compute
At the core of the upgraded Rivian autonomous driving platform is a proprietary processor called the Rivian Autonomy Processor, or RAP1. Designed entirely in-house, RAP1 powers the third-generation Autonomy Compute Module (ACM3), capable of processing nearly five billion pixels of sensor data every second. This level of compute is critical for interpreting complex driving environments at highway and urban speeds.
RAP1 also supports low-latency interconnect technology, enabling additional chips to be linked together to scale processing performance. This modular approach allows Rivian to future-proof its autonomy stack as algorithms, models, and sensing demands continue to grow.
Path toward higher levels of automated driving
The expanded compute capacity of ACM3 supports Rivian’s roadmap toward more advanced automated driving systems, including future SAE Level 4 capabilities. These systems are designed to operate without continuous human supervision under defined conditions, a milestone that depends heavily on both sensor fidelity and inference speed.
Rivian emphasized that robust perception remains the foundation of hands-free driving reliability. By fusing camera, radar, and lidar data, the platform can more accurately detect objects, predict motion, and respond to edge cases that are difficult for single-sensor systems to handle.
Software-first strategy and data-driven learning
Rivian detailed its software-first philosophy for autonomy, powered by the Rivian Autonomy Platform and an end-to-end data training loop. Fleet vehicles continuously collect real-world driving data, which is filtered through a structured data policy to identify rare and challenging scenarios. These scenarios are then used to train large-scale models offline, allowing the system to improve over time.
This closed-loop learning framework, often described as a data flywheel, enables faster iteration compared with traditional rule-based approaches. According to Rivian leadership, it is especially effective for handling complex situations where human-coded logic struggles to scale.
Large Driving Model enhances decision-making
A key component of the platform is Rivian’s Large Driving Model, trained on extensive datasets gathered from its vehicle fleet. The model uses Group-Relative Policy Optimization to evaluate multiple potential driving trajectories simultaneously. In complex environments such as busy intersections, the system compares different paths, selects the safest and most efficient option, and feeds that outcome back into the model.
When paired with Rivian’s multi-modal sensor suite, this approach improves situational awareness and enables more confident autonomous decision-making in real time.
Expanded hands-free driving and subscription plans
Rivian also announced near-term software updates for second-generation R1 vehicles. These updates include the rollout of Universal Hands-Free assisted driving across a significantly expanded road network. Coverage will increase to more than 3.5 million miles across the United States and Canada, compared with approximately 150,000 miles previously.
Gen 3 autonomy hardware timeline
Validation of Rivian’s Gen 3 autonomy hardware, including ACM3 and lidar integration, is currently underway. The company expects this hardware to debut with the R2 SUV toward the end of 2026. Rivian leadership described the upcoming platform as one of the most powerful combinations of sensing and inference compute planned for consumer vehicles in North America, reinforcing the company’s strategy of deep vertical integration across hardware and software.
Rivian confirmed that its Gen 2 vehicles and upcoming Gen 3 models will feature a significantly upgraded autonomous sensor and compute architecture. For the first time, the system combines high-resolution cameras, multiple radar units, and lidar, creating a multi-modal sensing approach designed to enhance real-time perception, redundancy, and operational safety.
Rivian autonomous driving platform built on in-house AI compute
At the core of the upgraded Rivian autonomous driving platform is a proprietary processor called the Rivian Autonomy Processor, or RAP1. Designed entirely in-house, RAP1 powers the third-generation Autonomy Compute Module (ACM3), capable of processing nearly five billion pixels of sensor data every second. This level of compute is critical for interpreting complex driving environments at highway and urban speeds.
RAP1 also supports low-latency interconnect technology, enabling additional chips to be linked together to scale processing performance. This modular approach allows Rivian to future-proof its autonomy stack as algorithms, models, and sensing demands continue to grow.
Path toward higher levels of automated driving
The expanded compute capacity of ACM3 supports Rivian’s roadmap toward more advanced automated driving systems, including future SAE Level 4 capabilities. These systems are designed to operate without continuous human supervision under defined conditions, a milestone that depends heavily on both sensor fidelity and inference speed.
Rivian emphasized that robust perception remains the foundation of hands-free driving reliability. By fusing camera, radar, and lidar data, the platform can more accurately detect objects, predict motion, and respond to edge cases that are difficult for single-sensor systems to handle.
Software-first strategy and data-driven learning
Rivian detailed its software-first philosophy for autonomy, powered by the Rivian Autonomy Platform and an end-to-end data training loop. Fleet vehicles continuously collect real-world driving data, which is filtered through a structured data policy to identify rare and challenging scenarios. These scenarios are then used to train large-scale models offline, allowing the system to improve over time.
This closed-loop learning framework, often described as a data flywheel, enables faster iteration compared with traditional rule-based approaches. According to Rivian leadership, it is especially effective for handling complex situations where human-coded logic struggles to scale.
Large Driving Model enhances decision-making
A key component of the platform is Rivian’s Large Driving Model, trained on extensive datasets gathered from its vehicle fleet. The model uses Group-Relative Policy Optimization to evaluate multiple potential driving trajectories simultaneously. In complex environments such as busy intersections, the system compares different paths, selects the safest and most efficient option, and feeds that outcome back into the model.
When paired with Rivian’s multi-modal sensor suite, this approach improves situational awareness and enables more confident autonomous decision-making in real time.
Expanded hands-free driving and subscription plans
Rivian also announced near-term software updates for second-generation R1 vehicles. These updates include the rollout of Universal Hands-Free assisted driving across a significantly expanded road network. Coverage will increase to more than 3.5 million miles across the United States and Canada, compared with approximately 150,000 miles previously.
- Looking ahead, Rivian plans to introduce a new autonomy-focused subscription offering called Autonomy+. The service will launch in early 2026, offering continuously improving features through software updates. Pricing is set at a one-time payment of $2,500 or a monthly subscription of $49.99.
Gen 3 autonomy hardware timeline
Validation of Rivian’s Gen 3 autonomy hardware, including ACM3 and lidar integration, is currently underway. The company expects this hardware to debut with the R2 SUV toward the end of 2026. Rivian leadership described the upcoming platform as one of the most powerful combinations of sensing and inference compute planned for consumer vehicles in North America, reinforcing the company’s strategy of deep vertical integration across hardware and software.
Press release
Click above to visit the official source.
Share: