It appears that Tesla’s internal beta of FSD Beta 10.13 was successful, as reports recently emerged that the highly-anticipated iteration of the advanced driver-assist system has started rolling out to consumer vehicles. A good portion of the update’s release notes has also been shared online.
As former Tesla AI Director Andrej Karpathy announced his departure from the company last week, CEO Elon Musk remarked that FSD Beta 10.13 should handle some tough left turns. Musk also remarked that the recent 10.x updates of FSD Beta are already “mostly at v11” in terms of their capabilities, though FSD Beta v11 itself should still be released hopefully next month.
Interestingly enough, one of the first observed vehicles that received FSD Beta 10.13 was a Model Y from Gigafactory Texas, a vehicle with a 4680 battery pack. This is quite an interesting development, as FSD Beta and 4680 battery packs are among Tesla’s most important projects in recent years.
Just as stated by Elon Musk, FSD Beta 10.13’s release notes directly reference beta tester Chuck Cook’s extremely challenging unprotected left turn, as well as other notable improvements on false positives and false slowdowns around crosswalks. While FSD Beta 10.13 would most definitely not be the final iteration of the advanced driver-assist system, it promises to be a substantial improvement overall.
Following are (partial) release notes of FSD Beta v10.13.
FSD Beta v10.13 Release Notes
- Improved decision making for unprotected left turns using better estimation of ego’s interaction with other objects through the maneuver.
- Improved stopping pose while yielding for crossing objects at “Chuck Cook style” unprotected left turns by utilizing the median safety regions.
- Made speed profile more comfortable when creeping for visibility, to allow for smoother stops when protecting for potentially occluded objects.
- Enabled creeping for visibility at any intersection where objects might cross ego’s path, regardless of presence of traffic controls.
- Improved lane position error by 5% and lane recall by 12%…
- Improved lane position error of crossing and merging lanes by 22% by adding long-range skip connections and a more powerful trunk to the network architecture.
- Improved pedestrian and bicyclist velocity error by 17%, especially when ego is making a turn, by improving the onboard trajectory estimation used as input to the neural network.
- Improved animal detection recall by 34% and decreased false positives by 8% by doubling the size of the auto-labeled training set.
- Improved detection recall of far away crossing vehicles by 4% by tuning the loss function used during training and improving label quality.
- Improved the “is parked” attribute for vehicles by 5% by adding 20% more examples to the training set.
- Upgraded the occupancy network to detect dynamic objects and improved performance by adding a video module, tuning the loss function, and adding 37k new clips to the training set.
- Reduced false slowdowns around crosswalks by better classification of pedestrians and bicyclists as not intending to interact with ego.
- Reduced false lane changes for cones or blockages by preferring gentle offsetting in-lane where appropriate.
- Improved in-lane positioning on wide residential roads.
- Improved object future path prediction in scenarios with high yaw rate.
- Improved speed limit sign accuracy on digital speed limits by 29%, on signs with difficult relevance by 23%, on 3-digit speeds by 39%, and on speed limit end signs by 62%. Neural network was trained with 84% more examples in the training set and with architectural changes which allocated more compute in the network head.
Don’t hesitate to contact us with news tips. Just send a message to simon@teslarati.com to give us a heads up.
News
Tesla supplier Samsung preps for AI5 production with latest move
According to a new report from Sedaily, Samsung is accelerating its preparation for U.S. production of the AI5 chips by hiring veteran engineers for its Customer Engineering team.
Tesla supplier Samsung is preparing to manufacture the AI5 chip, which will launch the company’s self-driving efforts even further, with its latest move.
According to a new report from Sedaily, Samsung is accelerating its preparation for U.S. production of the AI5 chips by hiring veteran engineers for its Customer Engineering team, which will help resolve complex foundry challenges, stabilize production and yields, and ensure manufacturing goes smoothly for the new project.
The hiring push signals that Tesla’s AI5 project is moving forward quickly at Samsung, which was one of two suppliers to win a contract order from the world’s leading EV maker.
🚨🚨 FIRST LOOK at Tesla’s AI5 chip, which will be available in late 2026 or early 2027 pic.twitter.com/aLomUuifhT
— TESLARATI (@Teslarati) November 6, 2025
TSMC is the other. TSMC is using its 3nm process, reportedly, while Samsung will do a 2nm as a litmus test for the process.
The different versions are due to the fact that “they translate designs to physical form differently,” CEO Elon Musk said recently. The goal is for the two to operate identically, obviously, which is a challenge.
Some might remember Apple’s A9 “Chipgate” saga, which found that the chips differed in performance because of different manufacturers.
The AI5 chip is Tesla’s next-generation hardware chip for its self-driving program, but it will also contribute to the Optimus program and other AI-driven features in both vehicles and other projects. Currently, Tesla utilizes AI4, formerly known as HW4 or Hardware 4, in its vehicles.
Tesla teases new AI5 chip that will revolutionize self-driving
AI5 is specialized for use by Tesla as it will work in conjunction with the company’s Neural Networks, focusing on real-time inference to make safe and logical decisions during operation.
Musk said it was an “amazing design” and an “immense jump” from Tesla’s current AI4 chip. It will be roughly 40 times faster, and have 8 times the raw compute, with 9 times the memory capacity. It is also expected to be three times as efficient per watt as AI4.
“We’re going to focus TSMC and Samsung, initially, on AI5. The AI5 chip, design by Tesla, it’s an amazing design. I’ve spent almost every weekend for the last few months with the chip team working on AI5.”
It will be 40x better than the AI4 chip, Musk says.
— TESLARATI (@Teslarati) October 22, 2025
AI5 will make its way into “maybe a small number of units” next year, Musk confirmed. However, it will not make its way to high-volume production until 2027. AI5 is not the last step, either, as Musk has already confirmed AI6 would likely enter production in mid-2028.
News
Tesla discloses interesting collaboration partner for Supercharging
This BOXABL collaboration would be a great way to add a rest stop to a rural Supercharging location, and could lead to more of these chargers across the U.S.
Tesla disclosed an interesting collaboration partner in an SEC filing, which looks like an indication of a potential project at Supercharger sites.
Tesla said on Tuesday in the filing that it was entering an agreement with BOXABL to design and build a Micromenity structure. Simply put, this is a modular building, usually a few hundred square feet in size, and it has been seen at Superchargers in Europe.
In Magnant, France, Tesla opened a small building at a Supercharger that is available to all EV owners. There are snacks and drinks inside, including ice cream, coffee, a gaming console, and restrooms. It gives people an opportunity to get up and out of their cars while charging.
This building was not built by BOXABL, but instead by bk World Lounges. It is likely the final Supercharging stop before people get to Paris, as it is located 250 kilometers, or 155 miles, from the City of Light.
Voir cette publication sur Instagram
Magnant has 56 stalls, so it is a large Supercharging stop compared to most. The building could be a sign of things to come, especially as Tesla has opened up larger Supercharger stations along major roadways.
It is for just a single building, as the Scope of Work within the filing states “a comprehensive package for one Micromenity building.”
NEWS: BOXABL, a company that creates modular, prefabricated buildings, has entered into an agreement with @Tesla.
This is Tesla formally contracting BOXABL to design, engineer, and build a pilot “Micromenity” structure, a compact, modular building unit.
While some info in the… pic.twitter.com/RabJczGpEp
— Sawyer Merritt (@SawyerMerritt) December 9, 2025
Superchargers are commonly located at gas stations, shopping centers, and other major points of interest. However, there are some stops that are isolated from retail or entertainment.
This BOXABL collaboration would be a great way to add a rest stop to a rural Supercharging location, and could lead to more of these chargers across the U.S.
Tesla has done a lot of really great things for Supercharging this year.
Along with widespread expansion, the company launched the “Charging Passport” this week, opened the largest Supercharger in the world in Lost Hills, California, with 168 chargers, opened the Tesla Diner, a drive-in movie restaurant in Los Angeles, and initiated access to the infrastructure to even more automakers.
Elon Musk
Tesla CEO Elon Musk confirms Robotaxi safety monitor removal in Austin: here’s when
Musk has made the claim about removing Safety Monitors from Tesla Robotaxi vehicles in Austin three times this year, once in September, once in October, and once in November.
Tesla CEO Elon Musk confirmed on Tuesday at the xAI Hackathon that the company would be removing Safety Monitors from Robotaxis in Austin in just three weeks.
This would meet Musk’s timeline from earlier this year, as he has said on several occasions that Tesla Robotaxis would have no supervision in Austin by the end of 2025.
On Tuesday, Musk said:
“Unsupervised is pretty much solved at this point. So there will be Tesla Robotaxis operating in Austin with no one in them. Not even anyone in the passenger seat in about three weeks.”
Musk has made the claim about removing Safety Monitors from Tesla Robotaxi vehicles in Austin three times this year, once in September, once in October, and once in November.
In September, he said:
“Should be no safety driver by end of year.”
The safety driver is just there for the first few months to be extra safe.
Should be no safety driver by end of year.
— Elon Musk (@elonmusk) September 4, 2025
On the Q3 Earnings Call in October, he said:
“We are expecting ot have no safety drivers in at least large parts of Austin by the end of this year.”
Finally, in November, he reiterated the timeline in a public statement at the Shareholder Meeting:
“I expect Robotaxis to operate without safety drivers in large parts of Austin this year.”
Currently, Tesla uses Safety Monitors in Austin in the passenger’s seat on local roads. They will sit in the driver’s seat for highway routes. In the Bay Area ride-hailing operation, there is always a Safety Monitor in the driver’s seat.
Three weeks would deliver on the end-of-year promise, cutting it close, beating it by just two days. However, it would be a tremendous leap forward in the Robotaxi program, and would shut the mouths of many skeptics who state the current iteration is no different than having an Uber.
Tesla has also expanded its Robotaxi fleet this year, but the company has not given exact figures. Once it expands its fleet, even more progress will be made in Tesla’s self-driving efforts.
Tesla supplier Samsung preps for AI5 production with latest move
Tesla discloses interesting collaboration partner for Supercharging
