Tesla Full Self-Driving (FSD) Explained: Features, Cost & Availability

Tesla says Full Self-Driving is headed toward unsupervised driving, but what's actually happening under the hood? A nearly 40-minute real-world test of FSD v13.2 in dense city traffic shows just how close the software is getting. We break down what it can do, what it costs, and how availability differs between the US and Germany.

Tesla FSD at a Glance: What the System Actually Promises

Full Self-Driving, or FSD, is Tesla's software package for what is currently partial automation with the potential to grow into something more advanced. The name is deliberately ambitious, since it suggests complete autonomy. In practice, though, FSD today is a driver-assistance system that falls under Level 2 on the widely used SAE scale for autonomous driving: the driver has to stay attentive and ready to take over at any moment, even though the car is steering, accelerating, and braking on its own. That puts FSD in a fundamentally different category from true autonomous systems such as robotaxis, where no human needs to intervene at all.

This distinction matters because FSD is treated differently depending on where you drive it. In the United States, Tesla has already begun operating unsupervised robotaxi service, without a safety driver, in a handful of cities. In Germany and the broader EU, FSD only received EU type approval in spring 2026, after the Dutch vehicle authority RDW completed the technical assessment and Germany's KBA adopted it; as of mid-2026 it is rolling out to German subscribers as FSD (Supervised), the same Level 2, ADAS-style assistance system used in the US, rather than the unsupervised, driverless function Tesla is now testing domestically. Europe's approval process has generally moved more slowly and incrementally than in the US, which is why the newest capabilities still tend to land there first. That gap is exactly why it is worth looking past the marketing language and examining how the software actually performs behind the wheel. One of the more revealing examples comes from AI DRIVR, a YouTube channel that specializes in Tesla FSD testing and documented version 13.2 during an unedited night drive through heavy city traffic.

Source: AI DRIVR: Tesla AI Delivered. FSD 13 is MIND BLOWING

FSD v13.2 Put to the Test: What the Night Drive Really Shows

The video was published in early December 2024 and quickly picked up by outlets like Electrek, InsideEVs, and Tesla Oracle, which lends real credibility to the footage. Rather than a polished demo run, the reviewer drives a genuinely demanding route at night, complete with intersections, construction zones, and heavy traffic. Conditions like these, uncontrolled and unpredictable, are exactly what matters when judging the real capability of the underlying AI and machine-learning models that power FSD. Over roughly 40 minutes, the route alternates between main roads, narrow side streets, and multi-lane intersections, giving a far broader picture than most short promotional clips from automakers.

Anticipatory Driving in Heavy Traffic

One key moment from the test: FSD recognizes on its own that a car two positions ahead is braking hard, and it reacts before the situation turns critical. Instead of slamming on the brakes, the system calmly changes lanes ahead of time to preserve a safe following distance. That kind of behavior requires continuously tracking several vehicles at once, not just the car directly in front, which says a lot about how far the system's sensor fusion has come. According to the reviewer, similar forward-looking behavior shows up around construction zones too, where the car merges well ahead of the narrowed lane rather than waiting until the last moment.

Complex Turns and Environmental Awareness

Maneuver chains are especially tough for driver-assistance systems, and one clear example is a right turn immediately followed by a left turn in dense traffic. The reviewer says v13.2 handles sequences like this noticeably more confidently than the previous version, FSD v12. In another scene, the system spots the risk of curb damage to a wheel or tire while turning left and actively steers right to avoid it before completing the maneuver. Details like this suggest the camera-based perception stack is factoring in more than just other road users; it's also picking up on fine details of the road infrastructure itself.

Everyday situations such as parking, backing out of a spot, reversing, and rolling over speed bumps also come off noticeably smoother and more comfortable in v13.2 than they did in FSD v12. Improvements like these often matter more to everyday users than a single dramatic maneuver, because they shape how much a driver trusts the system across an entire trip. Smooth, jerk-free behavior at low speed is frequently seen by testers as a more reliable sign of a system's maturity than any one flashy highway clip.

Weaknesses Documented Honestly

What stands out about this particular test is that the reviewer doesn't just praise the system, he also calls out the problems that remain. Those included:

• A manual takeover because the car crept into a restricted zone behind an ambulance
• Brief hesitation at stop signs in confusing situations
• Uncertainty when deciding how to handle a dead end
• A handful of rare situations with no clear rule to follow

Despite those isolated cases, the overall takeaway from the test is that the driving stayed calm and human-like throughout, even in unexpected situations, without jerky steering or braking. That kind of consistency across a full 40-minute drive is what many observers point to as evidence that unsupervised driving is getting technically closer than it seemed just a few years ago. Context still matters, though: a single unedited video is not a substitute for a systematic study, but it does offer a far more realistic impression than a short, cherry-picked marketing clip.

FSD Cost and Availability: US vs. Germany

Tesla used to offer FSD two ways, as a one-time purchase or as a monthly subscription that could be canceled anytime. As of February 2026, Tesla discontinued the one-time purchase option entirely, so FSD is now subscription-only, which actually works well for drivers who just want to try the features out, say, ahead of a long road trip. Exact pricing changes fairly often and varies by vehicle model and hardware generation, so it's worth checking Tesla's current price list before committing to a subscription. On top of the software cost, drivers still face the usual expenses for charging, insurance, and maintenance, which exist independently of whichever FSD package is chosen and shouldn't be left out of a realistic total-cost picture.

The Gap Between the US and Germany

In the US, Tesla has already moved from driver-assistance features toward actual unsupervised robotaxi service, now running without a safety driver in several cities. In Germany, FSD only cleared EU type approval in spring 2026 and is now rolling out as FSD (Supervised), the same driver-assistance category available in the US, rather than the unsupervised version Tesla operates domestically. Drivers using a Tesla in Germany get the same camera hardware and most of the same underlying software, but because European approval processes tend to move more incrementally, they can generally expect the newest capabilities, and any move toward unsupervised driving, to arrive there later than in the US. Charging infrastructure is a related piece of the puzzle worth factoring in, since autonomous and semi-autonomous driving features will increasingly shape how and where people charge; our EV fast charging guide is a useful starting point there.

Anyone weighing a Tesla purchase shouldn't look at FSD in isolation. It's one line item among several, alongside the vehicle price, insurance, and energy costs. Taking an honest look at the total cost of owning an EV helps put a software package like FSD into perspective, rather than treating it as a separate decision from the rest of the purchase.

Where Tesla FSD Is Headed Next

The version 13.2 shown in the test marks a real step up from earlier releases, largely because Tesla has been shifting the whole system toward a single, camera-based neural network that learns directly from massive volumes of driving data instead of relying on rigid, hand-coded rules. That approach sets Tesla apart from many competitors, who lean more heavily on LiDAR and radar sensing alongside cameras. Whether a pure camera-based approach is the more robust long-term bet is still debated among engineers, particularly in poor visibility or unusual weather.

Tesla's stated long-term goal is to evolve FSD into a genuinely unsupervised driving system, one that could eventually underpin its own robotaxi service. That would put Tesla in direct competition with companies already running commercial robotaxi operations today, Waymo chief among them. Comparing Tesla's approach with how Waymo's robotaxis work highlights two very different paths toward the same destination: Waymo relies on a dense sensor suite and geographically limited operating areas, while Tesla is betting on a more flexible but technically riskier camera-centric system designed, in theory, to work anywhere.

For drivers outside the US, the open question is when, and how much of, this capability will get regulatory approval locally. Until then, FSD remains a genuinely capable assistance system that still requires an attentive driver. There's also a broader pattern worth noting here: the computing power, sensor hardware, and energy demands that autonomous driving requires go a long way toward explaining why self-driving cars are almost always electric. Anyone following the trajectory of autonomous driving should really view it alongside the broader shift to EVs, since the two trends are reinforcing each other rather than developing separately.

Taken together, the FSD v13.2 test shows Tesla making real, measurable progress, especially in anticipatory driving and complex maneuvers in city traffic. At the same time, specific edge cases, like restricted zones or dead ends, remain a challenge. For drivers considering FSD today, the honest takeaway is that it's already a genuinely useful, comfortable assistance system, but it is not yet true autonomy, and it's worth being clear-eyed about that difference.

Additional Video

For a more critical, hands-on take, this real-world road test of Tesla FSD v13.2.9 balances the hype with an honest look at where the system still stumbles.

Source: Electric Chris [Chris Vanderstock] – Tesla FSD 13.2.9 Review: The Good, The Bad, and The Boring.