Tesla starts limited rollout of FSD v13.2, and early tests show impressive new capabilities

Tesla has officially started the limited release of its Full Self-Driving (FSD) v13.2 to a select group of testers, and early reports from these testers showcase an impressive leap in performance and capabilities.

FSD v13.2 is part of Tesla’s latest software update, version 2024.39.10, bringing with it a suite of enhancements designed to improve the system’s end-to-end driving network. These updates include the integration of 36 Hz full-resolution video inputs and a redesigned neural network architecture powered by Tesla’s Cortex cluster. With these advancements, the photon-to-control latency has been halved, offering quicker and more accurate decision-making.

However, the headline feature of FSD v13.2 is its new “park-to-park” functionality. A new “Start FSD (Supervised) button on the screen allows for end-to-end hands-free driving, as this button now gives owners the option to start from a parked state, and from there navigate complex routes, and park itself at the destination. This new feature alone should eliminate a lot of the disengagements needed in previous versions of the software.

First Impressions from Early Testers

Early testers have been sharing their experiences online, and the consensus so far has been overwhelmingly positive. Veteran FSD tester @DirtyTesla shared a video of a flawless first drive, where his Model Y started from his garage, navigated snow-covered dirt roads, managed intersections, and parked autonomously in a crowded lot.

The real-world capabilities of FSD v13.2 were put to the test in challenging scenarios, such as unprotected left turns and U-turns on busy streets by longtime tester Chuck Cook, who noted that the system handled these situations with a level of confidence and precision akin to that of an experienced human driver. One specific improvement noted by Cook is that the vehicle no longer comes to a complete stop behind the stop line. This has been common source of frustration among FSD users as this is often unpractical and not how human drivers handle these situations, even though they are required by law.

Other testers, such as @AIDRIVR, praised the smoothness of the driving experience, comparing it to professional ride-hailing services. AIDRIVR even shared a video showing FSD now backs into a Supercharger spot without any driver input. Testers also observed that the persistent steering jitter from previous versions has been eliminated, further enhancing the overall driving experience.

Who Got FSD v13.2 & What’s Next?

The initial rollout is limited, both in terms of the number of customers who received it, but also that it is limited to HW4/AI4 vehicles only. Unfortunately, there has been no updates provided by Tesla or and members of the Tesla AI team about when HW3/AI3 users can expect to receive this update. The most recent FSD roadmap from Tesla also provided no insights on a timeline for a HW3/AI3 release.

If testing continues without any major bugs found, we should expect a wider rollout of FSD v13.2 to more HW4/AI4 vehicles in the next week or two. If this happens, we should also see a release of FSD v13.2 in Canada later this month.

As for v13 itself, upcoming improvements mentioned in the release notes include using audio inputs for better handling of emergency vehicles, better navigation predictions, and support for complex destination options like pulling over onto the side of the road, into garages or parking in designated spots.

Here are the full 2024.39.10 and FSD v13.2 release notes.

2024.39.10

FSD (Supervised) V13.2

FSD (Supervised) v13 upgrades every part of the end-to-end driving network.

Includes:

  • 36 Hz, full-resolution AI4 video inputs
  • Native AI4 inputs and neural network architectures
  • 4.2x data scaling
  • 5x training compute scaling (enabled by the Cortex cluster)
  • Reduced photon-to-control latency by 2x
  • Speed Profiles on both City Streets and Highways
  • Start FSD (Supervised) from Park with the touch of a button
  • Integrated unpark, reverse, and park capabilities
  • Improved reward predictions for collision avoidance
  • Improved camera cleaning
  • Redesigned controller for smoother, more accurate tracking
  • Dynamic routing around road closures, which displays them along an affected route when they are detected by the fleet

Upcoming Improvements:

  • 3x model size scaling
  • 3x model context length scaling
  • Audio inputs for better handling of emergency vehicles
  • Improved reward predictions for navigation
  • Improvements to false braking and slower driving in parking lots
  • Support for destination options including pulling over, parking in a spot, driveway, or garage
  • Efficient representation of maps and navigation inputs
  • Improved handling of camera occlusions
Are you buying a Tesla? If you enjoy our content and we helped in your decision, use our referral link to get C$2,600/US$2,000 off your purchase.
Previous Article

NHTSA opens up probe over Piedmont Cybertruck crash that killed three

Next Article

Tesla 2024 Holiday Update is packed with new features

You might be interested in …