Tesla CEO Elon Musk detailed the changes that are expected to come with a massive Full Self-Driving (FSD) update, which is set to roll out sometime in September, he revealed earlier this week.

Tesla has been refining its FSD suite for years, but it has never been as good as it is now. The focus is to get the suite to a point where interventions are no longer needed and drivers simply become passengers, as they will not be responsible for paying attention to the road.

Elon Musk teases crazy new Tesla FSD model: here’s when it’s coming

That version of FSD will come eventually, but not next month. However, there are dramatic improvements that will come with next month’s FSD update that will roll out to the public, Musk said:

“The FSD software update next month will be a major step-change improvement for rare conditions.”

The FSD software update next month will be a major step-change improvement for rare conditions

— Elon Musk (@elonmusk) August 8, 2025

Additionally, he provided specific details on what would change, hinting that the need for a driver to pay attention will be “substantially reduced,” but there are some “complex intersections, heavy weather, or unusual events” that will still require drivers to assume responsibility for the car:

“This will substantially reduce the need for driver attention, but some complex intersections, heavy weather or unusual events will still require attention.”

This will substantially reduce the need for driver attention, but some complex intersections, heavy weather or unusual events will still require attention.

Note, the Austin robotaxi FSD build is ~6 months more advanced than what is available in cars in America and there are…

— Elon Musk (@elonmusk) August 8, 2025

We have been teased about these types of updates before, but usually they involve some kind of mention of FSD being ready for unsupervised driving “by the end of the year.” Musk did not mention that here.

There is also the fact that Tesla has another FSD build in Austin for the Robotaxi suite that is more advanced than what is available to the public. It has performed well, Musk says, making claims that there are times when it feels “eerily human.”

Tesla Q2 2025 vehicle safety report proves FSD makes driving almost 10X safer

The improvements in FSD capabilities in subsequent releases are usually very evident. As Tesla continues to refine the suite for the public, it gains more confidence and becomes smarter through the collection of data and the use of neural networks.

The only thing left to wait for is the release itself, and we are hopeful it will roll out to the public in September, as Musk says.

The specs of the upcoming Tesla Model Y L has appeared in new Chinese regulatory filings, revealing key specifications including a six-seat layout and an extended range of up to 751 kilometers. The variant is expected to launch later this year alongside a new long-range Model 3 variant rated at 830 kilometers.

The updates were listed on the China Ministry of Industry and Information Technology’s (MIIT) latest batch of new energy vehicle models that are eligible for vehicle purchase tax exemptions.

Model Y L to debut with larger battery, six-seat layout

Listed under the model code TSL6500BEVBA0, the Model Y L will feature dual motors producing 142 kW at the front and 198 kW at the rear. It will be powered by a 465-kilogram 82.0-kWh lithium-ion battery from LG Energy Solution, with a pack energy density of 176 Wh/kg, as noted in a CNEV Post report. The long-range crossover achieves 751 km on the lenient CLTC cycle, making it Tesla’s highest-range Model Y to date in China despite its curb weight of 2,088 kg.

The “L” designation is believed to refer to the vehicle’s larger size and seating configuration, as the new variant is listed with six seats. It builds on Tesla’s strategy to diversify offerings in the Model Y lineup, which currently includes both RWD and AWD five-seat versions.

Model 3+ breaks record with 830 km CLTC range

Alongside the Model Y L, Tesla China also registered a new rear-wheel-drive Model 3, which was designated with the model code TSL7000BEVBR1. The vehicle boasts either 800 or 830 km of range on the CLTC cycle, depending on its trim. This marks the highest range yet for any Tesla vehicle in China.

The variant will use a 448-kilogram, 78.4-kWh LG-supplied battery with an energy density of 175 Wh/kg and a peak motor output of 225 kW. The vehicle’s curb weight is listed at 1,760 kg. The model was previously identified in filings as “Model 3+,” hinting at a possible tier above the existing long-range variant, which tops out at 753 km CLTC.

Elon Musk has confirmed that Tesla is stepping back from its Dojo supercomputer initiative and focusing instead on the company’s in-house AI5 and AI6 chips. 

The update was initially reported by Bloomberg News, which cited people reportedly familiar with the matter.

Project Dojo Ends

As per Bloomberg, Peter Bannon, who led Project Dojo, will be departing Tesla following Elon Musk’s decision to shut down the initiative. The publication claimed that Tesla’s Project Dojo team has lost about 20 members recently, and the remaining members of the initiative would be reassigned to other data center and compute projects within Tesla. 

Dojo was expected to help Tesla train its Autopilot and Full Self-Driving systems. With this change, Tesla will reportedly be increasing its reliance on external partners for its training hardware. These partners reportedly include Nvidia, AMD, and Samsung, among other companies.

Musk Explains Dojo Shutdown

In a series of posts on X, Elon Musk explained that it doesn’t make much sense for Tesla to divide its resources to scale two different chip designs. And since the company is putting in a lot of effort in its AI5 and AI6 chips–which would be used not for training but in actual products like Optimus and the Cybercab–it was time for Project Dojo to take a bow.

“It doesn’t make sense for Tesla to divide its resources and scale two quite different AI chip designs. The Tesla AI5, AI6 and subsequent chips will be excellent for inference and at least pretty good for training. All effort is focused on that.

“In a supercomputer cluster, it would make sense to put many AI5/AI6 chips on a board, whether for inference or training, simply to reduce network cabling complexity & cost by a few orders of magnitude. One could call that Dojo 3, I suppose. The difference in real-world performance between AI4 and AI5 is far more than any chip version I’ve ever heard of by a lot. It’s real good,” Musk wrote in his post.