Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register
This site may earn commission on affiliate links.
new cars can't get 12.3.1 currently
1711391470952-png.1031944
Perhaps this means 12.3.1 will be quickly merged to the main vehicle software to be available to factory versions too. Even just considering the new vehicles that can run FSD Beta ([currently 11.4.9 on factory versions?] so exclude Cybertruck and new Model 3(?)), there's the practical step of camera calibration for a new vehicle, and unclear how those necessary miles (and time) will affect delivery. Perhaps this initiative will result in pre-calibration or faster calibration process? A FSD Beta demo/experience with another vehicle that is already set up with 12.3.1 could work but probably not the same as driving your new vehicle.

Even if this "hard requirement" isn't immediately feasible, hopefully this is a sign of even more focused attention across Tesla to get 12.x ready for wide release and ongoing improvements. (Hopefully without "too many cooks in the kitchen" / "mythical man month" type delays.)
 
Even if this "hard requirement" isn't immediately feasible, hopefully this is a sign of even more focused attention across Tesla to get 12.x ready for wide release and ongoing improvements. (Hopefully without "too many cooks in the kitchen" / "mythical man month" type delays.)
I guess in typical Tesla fashion - Elon shoots off an email and everyone concerned will be scrambling for weeks or even months to implement it.
 
You may be right but within 2 minutes of manually initiating defrosting I could see better and another minute later FSD became available again. FSD did display a message that the cameras had poor visibility. So the question is why? And in general as I stated previously my newer Model Y fogs up much more than my first Model Y. So if the cameras had visibility problems couldn't Tesla react sooner and initiate defrosting? My wife was just about to take a picture of the message which I've never seen before when the message went away.
Maybe they don’t have the camera area sealed and the camera defogger needs the other defogger to activate?
 
  • Like
Reactions: FSDtester#1
One detail I forgot to mention from my drive report with 12.3 was a short gravel road. I've not seen anyone else mention gravel roads, so some may find this information interesting.

We stopped at my Brother in law's farm yesterday. Their driveway is what most would call a short gravel road; it's several hundred yards long with a circle at the house. V11 and earlier was pretty much useless on this road. The nav data doesn't show this road at all, it's just a field according to the map. I engaged 12.3 at the beginning of the circle, pointing away from the road where the new route started. It successfully followed it around the circle, but edged right into a gravel parking pad. At this point it came face to face with grass, for which it stopped. If it had stayed on the road portion, it would have been fine. Anyways, I had to intervene to drive across the grass and back onto the gravel, at which point 12.3 took over and did fine driving down to the main road. The issue here was it wanted to turn right, but the route went to the left. There was another car straight across from me at a stop sign, patiently waiting for me to go. I'm not sure what would have happened, because I intervened at this point so as to not annoy the other driver any more.

I apologize for not mentioning these interventions in my drive report, I kind of forgot about them while writing the report as I classified the un-listed gravel road driving as separate from the rest of the drive in my head.
 
I don’t really mean to split hairs here, but do you think there is a difference between SEEING and PERCEIVING? I’m pretty sure it SEES stuff down the road and somewhat around, but I don’t think it is bringing it into perception or perspective till much later.. at speed, temporally it’s not MUCH later, but relative to what is most in front and around, the real REAL TIME interpretation and rules/logic/AI based re-action but bring what is 40 yards + down the road into perception in parallel, that’s what I’m saying I don’t think it’s doing. As a driver, I, at speed on a highway I’m looking in front, side and rear either constantly or every 200-300 m/s. But I’m ALSO looking many vehicles in front of the car I’m following, looking for brake lights to illuminate, congestion growing etc.. probably every 2-3 seconds, probably more like 1-2 seconds. I have already built an expectation model for what is to COME farther down the road assuming everything at THAT point on the down the highway position remains same. And then of course making adjustments for even farther down the road - or at least mental preparations for.

THAT is the type of farther down the road PERCEIVING that I don’t yet see any evidence, either in discussion or on the screen - is happening.
For humans there’s a difference. For FSD there should be no difference between the two. I guess one could argue that ‘perceiving’ for FSD is recognizing an object as something it needs to monitor or react to.
 
My car is in service today and I have loaner that came with 2023.44.30.14 (FSD 11.4.9). I just got notified that the loaner has a software update to 2023.44.30.30 (FSD 12.3.1). So, I'm installing it so I can play with it while my car sits in service!
Hopefully, you got the extra coverage on the loaner for wheel damage.

Bah dump dump..
 
Snow storm is rolling in here in Colorado so planning to do a bit of testing (just on 12.3, not 12.3.1 yet) shortly. I expect we'll have 2-3" of snow for testing.

With just a light covering of snow a bit ago initial results are a little sus. On a 50mph highway (2 lanes in my direction) the car just... drifted... off into the right lane (we were in the left lane).

No blinker. clearly not a lane change per se. Visualization showed accurate lane markers. Super weird! I guess this is a downside of a fully AI approach. I did grab a dashcam snippet if anyone really wants to see (wasn't really that exciting).

There were no cars anywhere nearby so no safety issue. I assume adjacent vehicles would help it pick a reasonable path.
We had a storm overnight and out of curiosity I turned on FSD for my drive home from work this morning. The roads had clear(ish) areas where people were driving and snow/slush on either side and in the middle. 12.3 tried to center in the lane rather than center in the wheel marks meaning it simply was driving in the snow and slush rather than in the clear parts of the road. I turned it off after a few hundred feet.

* note: I only did this on a straight section of road with no other traffic around and was (even more) ready to take over. I do not recommend using FSD in any sort of inclement weather where driving, road grip or visibility is affected. FSD is not designed for this and trusting it in these situations is foolhardy at best.

IMG_4194.jpeg
 
V12.3. Today I was not in a rush and let the car choose its speed on the 100 mile trip I've reported on. The speed limit was 55 mph, I set 58 as the max, and the car drove between 47 and 52 mph, slower than necessary. I avoided tapping the accelerator and just pulled over occasionally to let tailgaters pass.

I didn't notice as much speeding and slowing as on past drives.
 
  • Informative
Reactions: FSDtester#1
V12 in my commute today was definitely worse than V11. Just far too slow and it gets too close to the center line way too often. It felt like a drunk was driving at times with it swerving in the lane and alternating speeds. It's sign detection seems worse too. Definitely didn't recognize a few places where the speed limit increases from 25 to 35 and kept the lower limit (while then driving 22 mph)

It also seems way less efficient from a power consumption view. Harsher acceleration and braking used up a bit more power than usual.

And this is nitpicky but V12 stops way too far behind the line a stop lights which risks not trigger the weight sensor at some intersections.
 
Last edited:
12.3.1's speed control is more robotic, but overall it makes fewer disengagement-worthy mistakes vs 12.3. I think 12.3.1 would have a lower tendency to curb the rims/wheels

Also Tesla keeps tweaking the V11 highway stack, so 12.3.1 behaves differently vs 12.3, 3.1 tends to get in the rightmost lane slightly later than 12.3, so it causes a little more anxiety
 
V12 in my commute today was definitely worse than V11. Just far too slow and it gets too close to the center line way too often. It felt like a drunk was driving at times with it swerving in the lane and alternating speeds. It's sign detection seems worse too. Definitely didn't recognize a few places where the speed limit increases from 25 to 35 and kept the lower limit (while then driving 22 mph)

It also seems way less efficient from a power consumption view. Harsher acceleration and braking used up a bit more power than usual.

And this is nitpicky but V12 stops way too far behind the line a stop lights which risks not trigger the weight sensor at some intersections.
Concur
Concur
Concur
 
Last edited:
Are they giving loaners again ... earlier they were just giving Uber coupons.
In my experience: When they have loaners, they hand out loaners. When they don't have loaners, they hand out Uber coupons. And, as an added gefinky, if one's car is in for Something Serious, if there's any prioritization, the Something Serious person gets the loaner.

A lot depends upon the Service Center and how busy they are.