Tesla has begun deploying FSD V14.3.3 to its fleet, and the headline change is a 33 percent increase in the top speed of Actually Smart Summon. The feature, which allows a vehicle to navigate a parking lot autonomously and drive to its owner or a designated pin, now operates at a maximum of 8 miles per hour, up from the previous 6 mph cap. While a 2 mph bump might sound modest in isolation, the update represents a meaningful improvement in real-world usability, cutting the time required to cross a typical parking lot by several seconds. The speed increase is part of a broader software release, designated version 2026.14.6.6, which also merges the Spring 2026 feature set with the Full Self-Driving branch for the first time. This convergence signals a more unified development approach, bringing convenience updates and autonomous driving improvements under a single release cycle.
The evolution of Actually Smart Summon and its speed problem
Actually Smart Summon has been one of Tesla’s more controversial features since its launch in September 2024. The concept is straightforward: a driver uses the Tesla mobile app to command the car to exit a parking spot and navigate autonomously through a parking lot to a specified location, either where the owner is standing or a dropped pin on the map. In theory, it is a compelling convenience for rainy days, large retail lots, or situations where the driver is carrying heavy items. In practice, the feature has been hampered by a speed limit that made it frustratingly slow. At 6 mph, the car crept through parking lots at a pace that often made walking the same distance quicker. In busy commercial environments, the slow-moving Tesla would accumulate a tail of waiting vehicles, creating social pressure and occasionally awkward standoffs. Drivers also had to remain vigilant, keeping the app open and a finger hovering over the stop button, because the system has been known to get involved in minor collisions with gates, bollards, and adjacent vehicles. The speed constraint was not arbitrary; it reflected a cautious approach to a challenging domain. Parking lots are unstructured environments with unpredictable pedestrians, shopping carts, tight corridors, and vehicles reversing out of spots. Tesla needed to ensure the system could perceive and react to these hazards reliably before increasing velocity.
What 8 mph means in practical terms
The jump to 8 mph translates into a 33 percent increase in top speed, which changes the user experience materially. To put numbers on it, crossing a 200-foot parking lot at the old 6 mph limit took approximately 23 seconds. At the new 8 mph cap, the same journey takes about 17 seconds. That six-second reduction per summon session may not seem transformative, but it accumulates across every use and reduces the gap between summoning and simply walking. More importantly, the faster pace makes the feature feel less like a gimmick and more like a genuine time-saver. The speed increase does not stand alone. It builds on significant architectural work in the preceding FSD releases. Version 14.3 rewrote the AI compiler using MLIR, a compiler infrastructure that allowed Tesla to optimize neural network execution more aggressively, yielding a 20 percent faster reaction time across the board. Then, in version 14.3.2, Tesla unified the AI model that powers consumer FSD, its commercial Robotaxi fleet, and the Summon feature into a single architecture. That unification is the critical enabler. The parking lot maneuver now runs on the same neural network that handles highway driving, giving it access to a much richer understanding of the environment. The system can process obstacles, pedestrians, and other vehicles more quickly and with greater confidence. That confidence is what allows Tesla to raise the speed limit without, in its assessment, compromising safety.
AI4 exclusivity and the growing hardware divide
The 8 mph speed increase is available only to vehicles equipped with AI4 compute, also referred to as Hardware 4. This is Tesla’s current-generation autonomous driving computer, and it is the only hardware platform that supports FSD V14 in full. Owners of vehicles still running Hardware 3 will not see the same speed cap. This is consistent with Tesla’s broader strategy of funneling advanced FSD capabilities to AI4 first. The company has acknowledged the gap and has been promising a streamlined version of the software for HW3 vehicles, sometimes referred to as V14 Lite, which would also serve international markets where regulatory conditions differ. That release, however, remains months away. For owners who purchased Full Self-Driving capability on a Hardware 3 car, the situation is increasingly frustrating. They paid for a software package that was supposed to improve over time, but the most meaningful improvements are now gated behind hardware they do not have. Tesla has stated it plans to build dedicated retrofit factories to address the upgrade backlog, but that promise offers little comfort to customers who have been waiting for years. The divide creates a two-tier ownership experience that is becoming harder to ignore, especially as feature gaps widen with each new FSD release.
Regulatory clearance paves the way for faster operation
The decision to raise the speed cap also comes in the wake of a significant regulatory development. Approximately six weeks before this release, the National Highway Traffic Safety Administration closed its investigation into Actually Smart Summon. The probe had examined 159 documented incidents involving the feature. NHTSA found that across the entire history of the feature, there were zero injuries and zero fatalities. The incidents that did occur were limited to minor property damage, most commonly involving parking gates, adjacent vehicles, and bollards. During the investigation period, Tesla issued six over-the-air software updates that addressed the issues NHTSA had raised. The agency determined that these fixes were adequate and closed the investigation without issuing a recall. With that regulatory cloud lifted, Tesla has more room to push the feature’s capabilities. Raising the speed limit is the first tangible outcome of that cleared path. It signals that the company believes the system is now robust enough to operate at a higher velocity without introducing unacceptable risk.
Additional changes in FSD V14.3.3
The speed increase is not the only change in today’s update. FSD V14.3.3 introduces a live intervention-free streak counter, displayed on the main in-car screen. This counter tracks how many consecutive miles the driver has covered without disengaging FSD. The moment the driver takes over, the counter resets to zero. The feature is clearly designed to gamify the driving experience, encouraging users to avoid unnecessary interventions and let the system handle more driving scenarios. It adds a subtle behavioral incentive that could help Tesla gather cleaner data on true system performance, since interventions that occur for comfort or preference reasons rather than safety-critical reasons can muddy the metrics. The update also revises the mandatory disengagement feedback menu that Tesla introduced in version 14.3.2. The original menu offered four options for why the driver disengaged: Preference, Discomfort, Critical, and Navigation. The new release simplifies these to Navigation, Parking, Critical, and Other. The change suggests Tesla is refining how it categorizes disengagements to get more actionable data from the feedback loop.
The merger of Spring 2026 features with the FSD branch
Beyond the FSD-specific changes, software version 2026.14.6.6 represents a notable organizational shift. It merges the Spring 2026 software features with the FSD branch for the first time. This means that convenience features, user interface updates, and other non-driving improvements that were previously developed on a separate track are now being integrated into the same release stream as autonomous driving updates. For Tesla owners, this could mean a more synchronized update experience, where they receive a comprehensive set of improvements in a single installation rather than waiting for different branches to converge. It also indicates that Tesla is maturing its software development process, moving toward a more unified codebase that can serve both the driving and the infotainment sides of the vehicle with fewer fragmentation issues. This integration is likely to become more important as Tesla continues to expand its software offerings and as the line between driving features and cabin features becomes increasingly blurred.
Industry context and competitive landscape
Tesla’s Actually Smart Summon exists in a broader context of automated parking and valet features being developed across the automotive industry. Several manufacturers offer remote parking assist systems that allow a vehicle to park itself while the driver stands outside, but few have attempted to combine that with autonomous navigation through a parking lot to a remote from the driver. Mercedes-Benz has its Intelligent Park Pilot, BMW offers Remote Park Assist, and Ford has experimented with Automated Valet Parking, but none of these systems operate at the scale or with the over-the-air update cadence that Tesla brings. The ability to improve a feature like Actually Smart Summon through software updates after the vehicle has been sold is a structural advantage that traditional automakers are still working to replicate. Every OTA update that increases capability reinforces the value proposition of a Tesla vehicle in a way that hardware-locked features cannot match. However, the hardware exclusivity of the speed increase also highlights a limitation of the OTA model: software can only do so much without the underlying compute power to support it. The gap between Hardware 3 and Hardware 4 is becoming a defining issue for Tesla’s upgrade strategy, and how the company manages that transition will have long-term implications for customer trust and retention.
Observations on the practical utility of the update
The speed increase is a genuine improvement, but it does not fundamentally transform the usefulness of Actually Smart Summon. The feature remains situational. It still requires the driver to maintain visual line of sight in many jurisdictions, and the responsibility for avoiding obstacles ultimately rests with the human monitoring the app. The reduction in travel time makes the feature more practical for quick retrieval in moderate-sized lots, but not large lots. For a car parked on the far side of a sprawling retail center, the time savings are more noticeable. For a car parked a few rows away, the difference is marginal. The intervention-free streak counter is a welcome addition for enthusiasts who enjoy tracking their FSD performance, but it does not address the underlying desire for Tesla to release comprehensive fleet-wide intervention data. Such data would allow independent tracking of FSD progress and provide transparency into how the system is improving over time. Without that data, owners and analysts are left to rely on anecdotal experience and occasional release notes to gauge progress.
Looking at the broader trajectory of FSD development
The V14.3.3 release fits into a larger narrative of incremental but meaningful improvement in Tesla’s autonomous driving software. The unification of the AI model across consumer, robotaxi, and summon use cases is a strategic move that should accelerate development. Instead of maintaining separate models for different driving contexts, Tesla can now train a single large neural network that handles all of them, benefiting from the data generated by every mode of operation. The MLIR compiler rewrite in V14.3 was a foundational change that made subsequent improvements possible, and the speed increase in this release is one of the first user-facing benefits of that work. The speed cap on Actually Smart Summon was never a limitation of the electric motors or the steering system; it was a limitation of the perception and decision-making software. Raising it required confidence that the system could handle the dynamics of faster movement in confined spaces. That confidence came from the architectural changes in the preceding releases. This pattern of foundational update followed by visible improvement is likely to continue as Tesla moves through the V14 cycle and toward whatever comes next.
The hardware divide between AI4 and HW3 remains the most contentious aspect of this release. Owners who bought into the FSD promise on older hardware are watching the gap widen. The promise of a dedicated retrofit factory is a long-term solution to a problem that feels immediate to those affected. In the meantime, the V14.3.3 update offers a clear demonstration of what the software can do when paired with sufficient compute, and it serves as a reminder that the pace of improvement is not uniform across the fleet. For owners of AI4 vehicles, the faster summon speed is a welcome quality-of-life improvement. For the broader ecosystem of Tesla enthusiasts and investors, the release is another data point in the ongoing story of how Tesla is slowly, methodically building toward a fully autonomous driving system. The regulatory clearance from NHTSA, the unification of the AI model, and the integration of spring features with the FSD branch all point toward a development organization that is becoming more mature and more focused on execution.
