Tesla’s FSD V13 Pushes HW4 Hardware Capabilities; End of Line for HW3?

By Karan Singh
Not a Tesla App

As Tesla continues to push the capabilities of its hardware as it prepares for autonomy with FSD Unsupervised, there has been a distinct divergence in features between HW4-equipped vehicles on FSD V13, and HW3-equipped vehicles on FSD V12.6. 

There are definitely reasons for this, and Tesla hacker Greentheonly, over on X, went into great detail on some of the reasons why this divergence has continued and is likely to continue.

If you’re interested in some background, we recommend the following articles. Although they’re not necessary, you may enjoy reading them before diving right in if you enjoy this topic.

Back in the summer of 2024, we reported that Tesla’s HW4 had yet to reach its full potential, and that HW3 still had some compute to work with. However, FSD has undergone a rapid evolution - there are now fundamental differences driven by the hardware capabilities and sheer scale of the latest builds of FSD.

FSD V13’s Gargantuan Neural Networks

FSD V13 is a substantial leap for Tesla. It features substantially larger neural networks - the core brain that processes all the environmental data and makes decisions. Those networks are far larger in V13 than in V12, which results in a much smoother and human-like expereience.

Tesla’s FSD computers contain two nodes. According to Green’s analysis of FSD V13, Node B, which handles the actual end-to-end driving logic, tripled in size from 2.3 GB in V12 to a staggering 7.5 GB in V13. That’s a substantial increase, and one of the main reasons why HW3 can’t handle FSD V13. While Interestingly, Node A on HW4 remains similar in size to FSD V12, using up about 2.3 GB.

However, this massive size is also pushing the limits of HW4. While this isn’t necessarily a big problem immediately - at the end of the day, there is a limited amount of compute and ram available and model sizes will likely grow in the future. In fact, Tesla is planning to scale model sizes in the next FSD version and has already said they’ll need to optimize them, as context size is limited by on-board memory.

HW3 Limitations

Since HW3 can’t run the latest FSD models, where does this leave millions of Tesla vehicles equipped with the now-legacy FSD hardware? Are they hitting a dead end? Not exactly - Tesla has promised to replace HW3 computers with new units during the Q4 2024 Earnings Call, but only for those who have purchased FSD and only if Tesla is no longer able to update HW3 vehicles.

There’s no doubt Tesla hopes to achieve autonomy and then work backwards to make it compatible with HW3, but that’s unlikely to happen with the limited ram available on HW3.

Tesla’s AI team has continued to say that they’re still working on updating HW3 vehicles - but that they will be taking a backburner to updates for HW4 until the team can optimize the model efficiently enough to run on the smaller footprint. That means V12.6 isn’t just V13 “lite” - it is a distinctly different software branch tailored specifically for HW3’s capabilities.

Green's analysis showed that HW3 under V12.6 runs NNs totaling 1.2 GB on Node A and 3.1 GB on Node B, which is significantly smaller than the footprint of V13. Back in November, we discussed how Tesla is running a smaller and optimized FSD model on HW3, with the specific goal of achieving similar performance to V13. FSD V12.6 appears to be the current implementation of that smaller model strategy.

Green also found that of the total neural nets used for FSD, 135 are shared between the current V12.6 and FSD V13.2 releases, which is the result of Tesla’s AI team working to optimize and refine the V13 model for HW3. So some neural nets are shared between the two sets of hardware, while others are specific to HW3 or HW4.

AI4 in the Future

The fact that V13 is already stretching HW4’s limits makes us wonder about Tesla’s imminent release of AI5. Last year, Tesla mentioned that AI5 would be 18 months away - which is now about 8 months away, and potentially even sooner to help solve the autonomy problem.

At the same time, Tesla is looking to launch Unsupervised FSD in June in Austin, Texas, with presumably HW4, which would indicate that they’re confident in its capabilities.

Unless Tesla launches the Robotaxi network with AI5, we don’t believe AI5 will be a “must have” item. Undoubtedly, it will result in similar improvements to HW4 compared to HW3. However, the software needed to take advantage of the newer hardware could still be years away.

All eyes will be on Tesla’s Robotaxi network launch and the hardware required for it. For now, HW4 owners are in good hands with future updates are on the horizon, and we’re certainly excited for what’s next.

Musk Talks Robotaxi Details: Fleet Size, Teleoperators, Avoid Certain Intersections, Scaling and More

By Karan Singh
CNBC

In a live two-part interview with CNBC’s David Faber yesterday, Elon Musk laid out his ambitious near-term plans for Tesla’s Robotaxi Network and the broader rollout of Unsupervised FSD. Speaking from Giga Texas, Elon confirmed that Tesla will be launching its initial Robotaxi pilot service on Austin’s roads by the end of June 2025.

Tesla Already Testing Robotaxi With Safety Drivers

The upcoming Austin launch is the start of Tesla’s long-stated Robotaxi ambitions. After promising true FSD for years, it appears that Tesla is finally ready to debut it, albeit it’ll be closed to the public initially, and it will only be available on Tesla-owned vehicles for now.

Elon expressed confidence in that end-of-June timeline, revealing some of the extent they’ve gone to behind the scenes. Tesla has had test vehicles driving in Austin 24/7, currently with drivers in the cars. However, they’ve seen essentially no interventions required from the safety drivers during the testing phase, which is good news.

The initial deployment in Austin will be deliberately small to ensure that everything operates as expected.

"So we want to be very careful with the first introduction of unsupervised full self-driving, meaning that there’s the cars driving around with no one in it... Well, yes, and sometimes no one in it at all. Just going to pick someone up. So the car obviously has to be incredibly safe." - Elon Musk on CNBC

How the Robotaxi Network Will Operate

Musk said that Tesla intends to launch with just 10 Robotaxis for the first week and then slowly increase it to 20, 30, 40, and higher. All of that is to be done without a safety driver. Tesla will still have teleoperators supervising the vehicles remotely if they get stuck, but other than that, they will be in motion without an occupant in the driver’s seat.

The goal is to hit 1,000 active Robotaxis within a few months and then focus on expansion towards San Francisco, Los Angeles, and San Antonio. While Tesla could start with 1,000 or 10,000 - given that the initial vehicles are just Model Ys, Tesla wants to keep a tight leash on the fleet and make sure that both the safety and experience are up to snuff.

Geofenced and Avoid Certain Intersections

To further increase safety during the early phases, Elon confirmed that Tesla will geofence the Robotaxis' operational areas within Austin, limiting them to parts of the city where Tesla has the highest confidence in the system's performance. He stated they would avoid intersections unless the system is proven to handle them well, or route around them. While the goal is for the cars to operate without safety drivers, there will be rigorous remote monitoring of the fleet's activities. 

Looking at the broader regulatory picture, Elon acknowledged the current complexities, with rules varying by state and even by city. He reiterated his call for a "unified set of national regulations" for self-driving vehicles in the U.S. to ensure consistency and prevent a confusing patchwork of differing rules as vehicles travel across jurisdictions.

This seems to be a great idea, at least initially. Autonomous cars will be safer if they adhere to the same rules everywhere rather than having different rule sets based on the state. However, one issue is that pedestrians and human-driven vehicles may expect the vehicles to either perform or refrain from certain actions, such as not turning right at red lights, and having different rules for self-driving vehicles could cause confusion.

Adding Your Vehicle to the Robotaxi Network

During the interview, Elon painted a picture of a rapidly expanding fleet of privately owned Teslas capable of unsupervised FSD. He offered a bold prediction for the near future:

"My prediction is that probably by the end of next year [end of 2026] we will have probably hundreds of thousands, if not—... Over a million Teslas doing self-driving in the U.S. ... Unsupervised full self-driving, where you do not need to pay attention." - Elon Musk on CNBC

Tesla eventually plans to allow vehicle owners to add or subtract their cars from the Robotaxi fleet, creating a model he likened to a combination of Uber and Airbnb. This would allow owners to generate revenue when their vehicle isn’t being used, with Tesla receiving a portion of the revenue.

He also mentioned that Tesla owners will likely even make more from their cars participating in the fleet than their lease costs. Elon seemed confident in Tesla's ability to manage the logistics, stating, "Tesla has all the ingredients necessary to offer a vast self-driving fleet overnight."

Elon on Vision-Only and Licensing FSD

Elon also took the opportunity to emphasize that Tesla’s approach to autonomy, based on vision, is the ideal approach. The road system is designed for humans with eyes, so using neural nets with cameras is the ideal solution.

"Because the way that the road system is designed is for AI. It’s basically, I should say, it’s for intelligence, biological neural net, and eyes. That’s how the whole road system is designed. So what will actually work best for the road system is artificial intelligence, digital neural nets, and cameras... And what we found is that, when you have multiple sensors, they tend to get confused. So do you believe the camera or do you believe lidar? - Elon Musk on CNBC

While Tesla previously included radar in its vehicles, the disagreements between the radar and camera data led to uncertainty, prompting Tesla’s engineers to turn the radars off instead. The decision wasn’t about the expense, as per Elon’s comments, but rather on the reducing utility of other sensor types when camera vision becomes more accurate.

Elon also confirmed that Tesla is "very much open" to licensing its FSD technology to other major automakers and is currently in discussions with several. "The more we demonstrate the capability of self-driving, the more that they will want to license it, and we’re happy to help," he added. 

Optimus and Learning

While the interview primarily focused on Tesla’s ambitions for autonomy, Elon also touched on other key aspects of Tesla’s business and future. He reiterated that Optimus and autonomy will “overwhelmingly” dominate the future financial success of Tesla. At multiple Earnings Calls, Elon has said much the same, so it's not surprising to see him stick to this belief.

Elon provided more color on Optimus's development, noting that while achieving the goal of billions of humanoid robots is at least a decade away, even with insatiable demand. He reaffirmed that the target of producing a million robots by 2030 was still reasonable. The training process is evolving; currently, human operators in motion capture (MOCAP) suits perform basic tasks (picking up objects, opening doors, dancing) to bootstrap the robot's intelligence. 

The next threshold, Elon explained, is for Optimus to learn from watching videos, like YouTube "how-to" guides, enabling dramatic "task extensibility." Following that, "self-play," where robots learn by interacting with objects, like a child with toys, guided by a reward function, will be key.

Within five years, Tesla envisions Optimus being capable of household chores like doing dishes, walking the dog, and even proactively understanding and fulfilling user needs. This level of AI advancement requires immense compute. Elon confirmed Tesla's own Dojo training program, based in New York, is contributing, though Tesla will continue to buy many GPUs from Nvidia.

Watch the CNBC Interview

Tesla Demonstrates FSD in Australia, Says No Regulatory Barriers [VIDEO]

By Karan Singh
Not a Tesla App

Like Europeans, Australian Tesla owners have been eagerly awaiting the arrival of FSD for years, but it looks like there’s finally some good news coming. Thom Drew, Head of Tesla Australia, has confirmed that the company is actively working to bring FSD to Australian roads and, most importantly, sees no regulatory challenges blocking its path.

This update from Thom follows a video that was shared by Tesla’s AI account on X, where they showed off FSD successfully navigating the streets of Melbourne.

FSD in the Outback

Speaking to Australia’s News.com, Thom Drew provided some context behind Tesla’s Australian FSD efforts. He noted that Elon’s push and focus is the deployment of FSD globally. The engineering team is working across a lot of markets to actively push out FSD wherever possible.

Notably, Drew emphasized the favorable regulatory environment in Australia. He specifically noted that “There’s currently no blockers in Australia to releasing FSD, as we have in North America.” While he didn’t provide a specific timeline for launch, it was clear he was enthusiastic.

The lack of a regulatory blocker is great news for Australians, as Tesla’s FSD efforts in Europe have been plagued with regulatory issues. While the Australian market poses some new challenges as a right-hand drive market, at least Tesla is in control there.

Hook Turn

At the end of the video above, FSD tackles something unique to Australia - the Hook Turn. This is a relatively unique maneuver that’s specific to Melbourne’s Central Business, where vehicles turn right from the far-left lane.

For those familiar with Melbourne, that maneuver is notoriously tricky—and at times even unsettling—since even experienced human drivers can struggle with the unusual complexity of that turn. Having FSD successfully demonstrate its skill in navigating these local complexities and laws is an indication that Tesla’s data-gathering methods work well and are scalable to unique location-based rules and regulations.

What This Means for Australian Owners

Given that Australian FSD is in active development and that there’s a clear regulatory path for Supervised FSD in Australia, that’s a good sign of things to come. It seems fairly likely that Tesla will be able to release FSD in Australia by the end of 2025, if not sooner.

Last year, Tesla initially said that RHD markets would receive FSD in Q1/Q2. With Q2 coming to an end in about a month, Tesla may not be too far off the mark here.

It's important to remember that this refers to FSD (Supervised), meaning the driver must remain vigilant and ready to take control at all times, consistent with its behavior in North America and China. However, Australia's lack of regulatory blockers is a huge advantage for Tesla compared to other regions like Europe and even China, where Tesla needed to jump through hoops to get FSD approved.

If you’re in Australia, keep an eye out; FSD may be coming sooner than you think.

Latest Tesla Update

Confirmed by Elon

Take a look at features that Elon Musk has said will be coming soon.

More Tesla News

Tesla Videos

Latest Tesla Update

Confirmed by Elon

Take a look at features that Elon Musk has said will be coming soon.

Subscribe

Subscribe to our weekly newsletter