Crowdsourced data shows that Tesla FSD 12.5 is three to five times better than FSD (Full self driving supervised) 12.3.6.
The FSD 12.5 is showing 744 miles between critical disengagement and 111 miles between any disengagement.
The FSD critical disengagements improved 3.8 times from 194 miles with 12.3.6.
There was also a 3.8 times improvement from 29 miles to 111 miles for miles between any disengagement.
There was a 4.33 times improvement between city miles to critical disengagement.
The city miles to disengagement are better than the highway miles to disengagement. The highway miles is still using the Autopilot software stack. The FSD software will be unifying city and highway software which will substantially reduce disengagements. Tesla will only switch to FSD over Autopilot when FSD is clearly statistically improved over Autopilot Highway. There will also be no issues when handing off between systems after a unified software stack.
The disengagements are maps, potholes, speed and obstacles. There are almost no lane issues which was a main problem in 12.3.6 and earlier versions. The critical disengagements could only be some fraction of situations with other cars and obstacles at this point. The obstacles disengagement issue could be something more simple that is fixable with FSD 12.5.X.
I think merging highway with slightly improved city could enable FSD 12.5.X to get to 1500-2000 miles per critical disengagement. There could then be another 3-5X improvement with Tesla FSD 12.6.X and get to 5000-10000 miles.
The crowdsourced data is showing a person in Saskatchewan getting disengagements with 12.5.1. They had one critical disengagement with 12.5. There is less Tesla data for Saskatchewan roads. However, I drove for years in Saskatchewan. The roads do not have much traffic and are pretty simple in the cities and towns. There are no weather issues in the summer but there will be ice and snow during the winter. There are unsigned and unmarked rural roads.
Tesla FSD Tracker is crowdsourced driving data for Tesla FSD.
Brian Wang is a Futurist Thought Leader and a popular Science blogger with 1 million readers per month. His blog Nextbigfuture.com is ranked #1 Science News Blog. It covers many disruptive technology and trends including Space, Robotics, Artificial Intelligence, Medicine, Anti-aging Biotechnology, and Nanotechnology.
Known for identifying cutting edge technologies, he is currently a Co-Founder of a startup and fundraiser for high potential early-stage companies. He is the Head of Research for Allocations for deep technology investments and an Angel Investor at Space Angels.
A frequent speaker at corporations, he has been a TEDx speaker, a Singularity University speaker and guest at numerous interviews for radio and podcasts. He is open to public speaking and advising engagements.
How many critical disengagement can a robotaxi have? After spending a lot of time using large language models I just don’t see how Tesla’s not going to hallucinate at times. I just don’t understand why Tesla has to assume everything is All or nothing when it comes to this technology. Other companies are using mapping and lidar in order to add redundancy to their systems
Interesting detail. For 12.5, there are no interventions on the highway miles, so far. That’s 402 miles. If the tester would have driven as far on the highway as in in the city, the miles to intervention would have been roughly 140 miles..
Agree, Brian. But 12.5.1 looks pretty bad, at least for the first 21 entries. 12 miles between interventions? Shouldn’t 12.5.1 be an improvement over 12.5?
I just bought a 18′ Model 3, which has HW2.5, with FSD seeming to f-i-n-a-l-l-y be getting as good as a human, I’m starting to think I’ll be upgrading to HW3 in the coming months.