UPDATE: Tesla Full Self Driving Option Is Removed From Model 3, S And X

Tesla Model 3


Update 10/19/2018:

The Full Self Driving option, previously listed below Enhanced Autopilot, has been removed on the Tesla Model S and Model X design studio as well. At the time the original article was published, the option had only been removed from the Model 3. According to Elon Musk, the option will still be available for about a week “off the menu”. 

Tesla Design Studio 10/19/2018 with full self driving option removed

Original article published 10/18/2018 included below:

Tesla Model 3, Model S and Model X FSD option is currently off the menu for “causing too much confusion”

Previously, buyers of the Tesla Model 3 could buy in to a future full self driving option. Reddit users spotted the option was missing earlier today after the announcement of the new mid-range Model 3.

Tesla FSD is missing

Now Elon Musk has addressed the issue. It seems that the option is still available “off menu” for the next week or so. Then will be going away for the time being.

Elon responds to Hoff

Elon Musk just as recently as a few days ago noted that buyers who opted for the full self driving option would be receiving hardware updates. This naturally led many to believe that the previously alluded to new custom chip might be on the way soon.

Last month, an additional 100 Tesla employees also signed up for testing fully autonomous driving.

Considering the timing of these previous announcements, the removal of the Model 3 Full Self Driving option might mean that the new hardware change could happen soon. If the hardware change is imminent, it would make sense to remove the option until it is rolled out.

Source: TwitterReddit

Hat tip to Dan Zorrilla for the heads up!


Categories: Tesla

Tags: , , ,

Leave a Reply

54 Comments on "UPDATE: Tesla Full Self Driving Option Is Removed From Model 3, S And X"

newest oldest most voted

Confusing article. One change is the impending elimination of the RWD long range Model 3 which Musk addressed. The other change is the elimination of the full self-driving option which Musk did not address in the Tweets shown in this article.

The response was to @TheHoff525 (aka Nick on twitter) and in response to the full self driving option.

I included a screen capture instead of the embedded tweet. The full twitter interface shows who he was responding to. 🙂 Link to the original tweet is now listed as a source instead.

I have been wondering if the shakeup in Tesla’s leadership would lead to the FSD option being removed. Very clearly offering that was premature, and it would make sense for the option to be removed until Tesla is much, much closer to reliable Level 4 autonomy than it is.

Of course it’s just speculation at this point to suggest that this “temporary” change will become much more than just a one-week suspension; to suggest that the FSD option will disappear for many months or perhaps a few years.

Again, my working theory is that Tesla will at some point relent and start putting either lidar or phased-array radar scanners into its cars. I think Tesla’s advance towards FSD will continue to be stalled so long as they stubbornly stick to the “we can get along with just cameras and a few low-res scanners” approach.

If I’m right, then Tesla will either have to offer a refund (with interest/penalty) to all those who have paid for FSD, or else Tesla will have to do a rather expensive retrofit to add active sensors to the cars of those who have previously paid for FSD.

Agree. It was a waste of money.

They’ll offer a 5k credit toward a new Tesla.

Is Tesla network still a go?

Most likely it is delayed until hardware is ready so Tesla doesn’t have to give you free hardware update. I would have paid for the feature, but now I will likely have to pay a lot more for the feature when it actually comes out.

Do Not Read Between The Lines

I think they’re taking it away from the Model 3 so that they can optimize production, not include updated hardware and not have to think about retrofits.

Right – this makes a lot of sense. Previously, people ordering FSD would possibly be getting a computer upgrade, if it was necessary. Now that Tesla knows they’ll definitely have to give them all computer upgrades, the option costs Tesla a lot more than they were anticipating to continue offering it.

So they’re not offering it for the next 6 months while they continue to work on the new CPU. People who order Model 3 between now and when the new CPU is standard will probably have to pay something like $10K to have their computer replaced as a retrofit.

Having lived with Autopilot for about a month now in my model 3 and being pretty familiar with the capabilities of modern digital cameras I am very skeptical Tesla is anywhere close to level 4/5 autonomy. I certainly find Autopilot a useful driving aid and love driving my model 3 which is a good thing because I don’t expect it to be driving me any time soon. If Tesla owners want to see for themselves the challenges of relying on current camera tech for providing sufficient data for self driving Tesla’s new dash cam feature will allow that. Just look at some of the footage from night time driving or driving in very high contrast scenarios. You will see a lot of missing data.

I don’t even get why any reasonable person would buy the assertion that you can use cameras as the primary sensors for self-driving cars. Normal video cameras can’t “see” at night any better than the human eye can; why cripple the sensors of self-driving cars in the same manner? Active sensors such as lidar and radar don’t care if it’s day or night outside, and obviously would be preferable for night-time driving.

Now, there is the possibility of using infrared cameras, but those require special and somewhat bulky cooling systems, and they would have to use different image recognition software, because the appearance of things in infrared images is different from their appearance in visible light images. Using visible light cameras in the day and infrared cameras at night would be very much against the KISS principle (Keep It Simple, Stupid).

Even infrared isn’t really a solution. In a totally dark room or a dark night things are usually black in infrared too. Security cameras rely on infrared LEDs to light up the environment. If you turn them off and have no other lights then all you get is a black screen.

Maybe they could have super bright IR spotlights shining out from the car. Unfortunately although that wouldn’t blind humans it seems like it would blind any other car with infrared sensors, so it’s not really a viable alternative.

That actually leads to another question: when you have tons of autonomous or semi-autonomous cars on the road, how much of a problem does interference become? If everyone is employing LIDAR and sonic sensors, what happens? I guess one plus with cameras is they just take in the light from the environment, they don’t actually send anything out.

Typically, headlights are on in nighttime driving.

Typically, headlights are aimed in only one narrow direction, and if the road curves, it’s not even the right direction. To avoid T-bone and rear-end collisions, we’ll want autonomous cars to continually scan 360° around the car. Do you envision mounting dozens of headlights all around the car?

Also, there is a real safety issue with cars overrunning their headlights. This is something which unfortunately auto makers have been slow to address.

I am a strong Tesla fan, but stubbornly insisting on cameras as the primary sensors for self-driving cars is — to be blunt — stupid. It’s doubly stupid if driving at night.

“…when you have tons of autonomous or semi-autonomous cars on the road, how much of a problem does interference become? If everyone is employing LIDAR and sonic sensors, what happens?”

This is a potential problem, and I’m surprised I haven’t seen the issue raised before in any InsideEVs discussion.

My guess is any long-range active sensors (lidar or radar) will use digitally pulsed beams in a specific pattern that is unique to each car. Cars will ignore scanning beams from other cars in the same fashion that your cell phone ignores calls to other cell phones.

But that’s just a guess; I have not seen any article which addresses the problem.

Ultrasonics would be far less of a problem. Those are such short-ranged sensors that there shouldn’t be much if any overlap from other vehicles. But I suppose even there, autonomous car designers would want to put in some sort of digital variance (frequency shifts, or pulses) to ensure one car isn’t “hearing” the ultrasonic signals from another.

I consider myself reasonable, and I think of it this way: We have ~30k automobile related deaths per year in the US, how many would we have if every driver was as good as the best driver in the world? I don’t know who the best driver in the world is, but I do know that he or she only has two eyeballs that can only look in one direction.

I absolutely want level 4/5 for the reasons you stated. I just think it is further away than many people think. And I do think Autopilot and other L2 features *when used properly* can improve safety. For example, on the interstate, having Autopilot deal with lane centering and watching the car directly ahead allows me to look to the sides more for hazards like deer, someone about to merge at a silly slow speed etc. But, Autopilot will also happily pace another car in their blind spot and take no action to avoid road debris. For example, if there is a paper bag in the road every driver knows this isn’t a big deal. A softball sized rock on the other hand could cause an accident if hit. Autopilot doesn’t consider or probably even see let alone evaluate either.

but he turn his head and has a lot of mirrors

“I don’t even get why any reasonable person would buy the assertion that you can use cameras as the primary sensors for self-driving cars”

That’s because you have a colossal ego. You know little about machine vision compared to those actually developing the technology, yet you talk like an expert.

If you’re actually interested in discussing the subject, instead of merely displaying that you know almost nothing about it, then you are invited to participate in this discussion thread:


BTW — I don’t need to be an “expert” on a subject to know more about it than many do.

Yes, you absolutely do need to be an expert for there to be even the slightest legitimacy in your opinion that Tesla’s camera+radar approach is “stupid”.

I never said that you don’t know more than “many”, whatever that means. I said you know little compared to those developing the technology. I’m a talented hardware/software engineer, but also know little compared to these leaders.

I’m fully open to the possibility that Tesla’s bet will backfire, but I also think it’s possible that it’s the right move. I don’t know, and cannot call either approach stupid or moronic with any conviction. Even if Tesla gets their stuff working well two years after LIDAR-based autonomy is first sold in mainstream cars, it would still be the smart move over the long run.

In the early 2000s, fuel cell cars seemed far more promising than EVs, with working prototypes having 400km range, and nobody thought batteries would ever reach $100/kWh without a radical discovery. It’ll be a few years before we find out whether Elon’s camera bet was wrong.

If HW3 is coming, I don’t want to sell you FSD today. It forces me to put HW2 in your car and then later I have to give you HW3 for free.

Why free, they were charging more for it, and they need money now vs into the future

Some income now at the expense of having to pay a lot more later… is short-term thinking, and it is the sort of management decision which has caused many formerly sound American companies to go bankrupt.

Thank goodness Tesla, as a general rule, doesn’t fall into that trap! I have some serious concerns about Elon Musk’s management style at Tesla, but I certainly do admire his unswerving dedication to long-term planning, and his rejection of the Wall Street “greed is good” philosophy which focuses on short-term gains at the expense of long-term financial health.

Current Tesla cars on the road will be too old before Tesla achieves level 4 self driving.

Good analysis, even if I hope you are wrong because it is going to cost Tesla a lot of money, whether it is refund or retrofit. Does anyone know how much they have earned on FSD so far?

There is no new leadership. Not yet anyway, this is all Elon Musk.

Honestly having an option of a future hypothetical feature never made sense.

I think it was based off the belief they were much closer to self driving than they really were. Tens or hundreds of millions have been spent on this by companies with many great engineers. It’s simply not an easy task to solve. We’re still years away from true autonomous driving, and it might even be decades.

Elon probably realized that his lidar free approach to FSD was moronic.

LIDAR-free autonomous driving will definitely happen in the next decade. It won’t be perfect, but it’ll be far better than the average driver.

It’s an open question whether that’ll happen before LIDAR systems solve their limitations (including weather issues, cost, and appearance), but it will happen. Such a prediction is hardly moronic. GM’s Supercruise uses LIDAR-gathered road data, but no LIDAR sensor in the cars.

…and Cadillac Super Cruise, like Tesla AutoSteer, is merely an automated lane-keeping function. Cadillac’s ABS plus Super Cruise, just like Tesla ABS plus Autopilot/Autosteer, cannot detect stationary obstacles when traveling at highway speed.

Both systems are Level 2 or at best 2+ autonomy, and what we need is reliable Level 4 autonomy.

More discussion here:


…while GM is trying to figure out how to get down Lidar cost from $20k to $300 while increasing range with 250%. good luck with that….

Solid state LIdar are now available for less than$500

Another Euro point of view

Likely your are right.

“Again, my working theory is that Tesla will at some point relent and start putting either lidar or phased-array radar scanners into its cars.”

WTFYTA they already have radar in the front of the car.

Semi-autonomous vehicles, including Tesla cars with ABS or Autopilot, have very low-res Doppler radar, which can’t detect stationary obstacles. That is orders of magnitude less useful than phased-array radar. That’s why semi-autonomous cars keep running into vehicles (such as fire trucks) parked on the highway.

More info here:


And see discussion here:


“Very clearly offering that was premature,…”

FUD from PP.

FSD release is imminent thanks to those neural super networks and secret chips.

Tesla is years ahead.

Go Elon!


Thank you for exposing how intellectually bankrupt all your serial Tesla bashing FÜD is. When there is a serious discussion of a real deficiency in Tesla’s cars, you have nothing at all to add to the discussion.

Begone, insect.

More like they are in deep legal or gov trouble about this scam

I wouldn’t say deep trouble, but I could see the SEC or FTC pressuring them.

Selling the FSD option is no different from selling anything on Kickstarter. There’s no legal issues given that Tesla isn’t breaking any promises regarding delivery of the product. Ethically, this purchase option has been one of the most questionable things Tesla has done, so this is likely about PR trouble.

Tomorrow is the two year anniversary of Tesla announcing HW2 and offering the FSD option.

They’re getting ahead of the impending PR disaster. People are still going to write about Tesla failing to deliver FSD within two years, and rightfully so, but by pulling the option now, at least Tesla won’t be removing the FSD option in reaction to the bad press.

I will reiterate my not-Tesla-software idea that FSD is not attainable as an upgrade with existing Tesla hardware. Even Tesla is saying that. They already backpedal on this point. The computer (and what else?) needs a hardware upgrade. We’ll see.

There is no “what else”. Eversince they came out with HW 2.0 they said it may not have enough horsepower for FSD, but the computer has been designed for easy upgradeability. They are 6 months out from introducing HW3 and FSD is not released nor approved by regulators yet.

There is very definitely a “what else” if Tesla needs to use lidar and/or phased-array radar, rather than relying on cameras as Elon keeps insisting.

I guess piling up people who want the HW3 upgrade they already paid for when available (I paid for it, give it to me naugh!!!!) would be a pain in the back…

Not sure what “off menu” ordering means (same comment for LR RWD)…

Off menu means it must be special ordered essentially.

The phrase is commonly used to refer to restaurants that have discontinued or special dishes that are not advertised or listed in the menu. But if a customer knows about the secret dish, they can order it.

Yes, that’s obvious.., but what does it mean in the context of ordering the RWD LR… eg ordereable in person only at a Tesla store? Or some direct contact process?

For example, a celebrity who is willing to pay an exorbitant price to get a custom paint color on his car, or custom interior accents inside.

Just means that FSD requires HW4+ which isn’t ready yet. And upgrade in the future would cost Tesla too much money.

Guess there is non point in selling obsolete hardware that will need a future retrofit with next gen hardware right around the corner.

“was causing too much confusion” = “we were full of sh1t and were gonna get our butts sued off of us if we kept charging people for vaporware”

(⌐■_■) Trollnonymous

I wouldn’t pat for that feature anyway so……….meh.

I don’t know if it changes anything but FSD is available to me for purchase on my already delivered M3.

But i was saying all along that FSD is much more difficult problem in terms of “last mile” (i.e., acceptable error rates) than many think — and not necessarily solely because of sensor types etc.

How Musk got fooled into this? well, you may have seen the autonomous flight, e.g., skydio.com. These things do exist. The problem is that there is a decade of research between 5% and 3% error rate. And 3% error rate is not nearly enough when we speak of human casualties. IMO Musk was ill-advised on this, or, which is more probably, was foolishly over-optimistic in his extrapolations. Learn some calculus, run some lab experiments for a couple of years, and you will be much more properly positioned to make such extrapolations.