V15의 릴리즈 노트에는 FSD (Supervised) v15.0.1이라고 나오는데 소개 동영상에는 Tesla Self-Driving V15라고 나오네요.
혹시 FSD(Full Self-Driving)라는 이름을 TSD(Tesla Self-Driving)로 바꾸려는 걸까요? 릴리즈 노트의 이름은 지금까지의 관성 때문에 붙은 것일 뿐이고요.
만약 이렇게 된다면 앞으로는 Supervised나 Unsupervised는 붙이지 않아도 될 것 같아 보이기도 하고
EU와 그 외의 DCAS 02 채택 국가들에 배포되는 것의 이름은 TAD(Tesla Assisted Driving)인데 북미와 호주/뉴질랜드 그리고 한국의 미국산 차량에 배포되는 것의 이름이 TSD이면 완벽한 조합이 될 듯합니다.
그리고 이렇게 되면 Full Self-Driving이라는 논쟁적인 이름을 Autopilot과 함께 역사의 뒤안길로 떠나보내면서 Unsupervised의 애매한 부분도 깔끔하게 해소할 수 있을 듯하고요.
한편 HW(SI)3용 FSD v14 Lite의 이름은 TSD v14.x Lite로 바뀔 듯하고 EU 등에 배포될 것의 이름은 TAD v14.x Lite가 될 듯합니다.
이상은 아무 근거 없는 저만의 개인적인 예측과 상상이었고, 제 미국산 2020년식 모델 3는 FSD v14.3 Lite 정도에 만족하면서 계속 타고 내년이나 후년 정도에 사려고 하는 모델 Y는 꼭 TSD(Tesla Self-Driving) V15나 V16이 되는 미국산 모델 Y로 사야겠다고 다짐하게 되네요. ^^
Tesla FSD V15 Offers Stunning Visualizations That Distract From “25x”
The cool visualization distracts from the much more important question of safety. If it's really 25 times safer, how bad was v14 which people praised so much?
https://t.co/oNLNEvuTo3
Full release notes below
– – –
v15 upgrades every part of the end-to-end driving network. Includes:
Upgraded vision encoder shared between FSD, Robotaxi, and Optimus, which is trained on millions of internet videos for improved generalization and data efficiency, improving the critical intervention rate by 25x.
Introduced Tesla Reality View, replacing all of the explicit model signals into a single end-to-end visualization using less total compute.
Improved navigation by using an end-to-end learned model based on route preference data from the fleet.
Improved parking location selection by using personalized parking history, and the map will now display multiple parking options at your destination.
Improved driver monitoring by increasing vision-based attention monitoring availability with hats and sunglasses and decreasing strictness in higher speed driver profiles.
Improved detection of road signs, small animals, road debris, and chains.
Improved detection of potholes, road details, and avoidance of deep standing water.
Improved speed handling for each driver profile, school zones and speed limits, and inclement weather.
Improved Autoshift functionality by unifying with FSD-based gear predictions.
FSD V12 Introduced end-to-end neural-network driving.
FSD V13 built on that and reduced interventions.
FSD V14 brought point-to-point travel and a huge leap in capability, helping FSD start to go mainstream.
FSD V15 will bring:
• Pothole avoidance
• AI-based mapping and navigation rather than relying on conventional third-party maps.
• A new user interface that will re-render reality in a "mind-bending way."
• Fleet intelligence that incorporates driving information collected across Tesla vehicles.
• Voice integration, allowing drivers to request different routes conversationally (likely through Grok).
• Improved parking, destination handling and trip planning.
• 10x parameter count increase
• New lane change and exiting behavior
• And eventually (2027 for Model Y), FSD Unsupervised for personal cars (Tesla's robotaxis already run V15, and the personal Cybercabs being delivered next month likely will too).
Some thoughts and expectations about FSD V15 put together by @grok
FSD v15 is reportedly going for “wide release” next month (December 2026) with the Holiday Update. Elon Musk stated it “goes to wide release next month” and “will be something special.”
It is HW4/AI4-only (HW3 stays on v14 Lite & uncertain if V15 can ever be ported to HW3).
V15 is purported to be a major architectural jump: ~10 billion parameters (roughly 10× the current ~1B model), earlier hazard prediction, faster reactions, stronger collision avoidance per @aelluswamy
It should include @TeslaSI own AI-based maps starting with trip start/end, a new “mind-bending” UI that re-renders reality, and reportedly has fixes for pothole avoidance, garage/home parking, HOV/carpool lane usage, and last-second lane changes for turns/merges. How much of these features will be included in the early V15 releases is unknown.
Major version jumps in FSD (v12→v13, v13→v14) have consistently introduced regressions alongside improvements. Expect the same with v15 along with many “dot” releases along the way.
What we may expect (or hope) to see with V15:
Addressing these common V14 pain points:
• Speed profiles too coarse (no fine max-speed offset); school zones & limit changes ignored & illogical or inappropriate speed selection
• Last-second exits/merges/turns + HOV refusal (Elon confirmed both “should be” fixed in v15)
• Brake stabbing, hesitation, phantom braking, swerves for skid marks/leaves
• Weak parking performance and selection (esp. garage/home) & pothole avoidance (both reportedly coming at “some point”)
• Difficulty detecting & reacting to thin chains, pivoting traffic arm barricades, & high obstacles such as lumber extending off the end of a loaded truck ahead
• Illogical routing, navigation database to reality mismatches, conditional sign misses, roundabout signaling
v15 expected gains: ~10B-param model (10×), earlier hazard prediction, faster reactions, better collision avoidance, Tesla AI maps (start/end of trips), new “mind-bending” UI.
Likely early v15 quirks:
• Higher initial interventions until fleet data refines it
• Over-cautious swerves or delayed moves from bigger model
• AI-map transition oddities near start/end
• UI visual overload at first
• Occasional HW4 latency in dense traffic
• Speed-profile friction remains; complex parking can temporarily regress
Major jumps always bring both big gains and early regressions. First builds will show the balance.
Personally, I really hope we do see a massive leap forward, but if it is a big jump in capabilities, it may take some time to get everything working smoothly. I also hope we get some clarity from @Tesla for the path forward for HW3 FSD owners and users.
@TeslaSI@elonmusk@aelluswamy@pduan
Now that the hype train has stared about FSD vXV being released with the December holiday update. Can you give any insight to the loyal HW3 kickstarter group that got you here?
Will there be a FSD vXV Lite? If so how long do you think it might take to distill down? If not, is v14 Lite the end of the road for HW3?
재미있는 상상. ^^
만약 @MKBHD가 오스틴을 배송지로 지정하고 주문한 Cybercab을 받아서 필요한 확안을 한 후 내기에 졌다고 인정하고 삭발을 했는데, 나중에 NHTSA가 테슬라의 FMVSS 자기인증은 무효라고 결론을 내면 어떻게 되는 건가요?
12월말까지 NHTSA가 결론을 내지 않으면 그는 NHTSA가 결론을 낼 때까지 삭발을 유예해야 할까요?
The Cybercab that MKBHD bet his hair on is due to reach buyers in December. The public still can't see Tesla's proof that it's legal to sell.
In September, regulators ordered Tesla to prove under oath that a car with no steering wheel and no pedals meets federal safety rules. One of those rules says the brakes must work from a foot pedal.
Zoox needed a federal exemption to run its robotaxi, capped at 2,500 a year. Tesla is selling the Cybercab without one.
Tesla's cover letter with its answers was one line:
"The entirety of this file is confidential business information."
Should Tesla have to show its answers before the first Cybercab reaches a buyer?
FSD v15가 나와도 당장 공식 비감독형을 표방할 가능성은 낮아 보입니다.
주행 책임이 제조사로 넘어가는 순간 리스크 프리미엄이 가격에 반영될 수밖에 없기 때문입니다. 로보택시 요금에 보험료가 녹아 있듯, 개인용 비감독 FSD 역시 보험 비용이 얹혀 지금보다 이용 단가가 크게 오를 것이고, 까다로운 법적 제약도 뒤따르게 됩니다.
결국 v15의 현실적인 포지션은 '비감독형에 준하는 감독형'일 가능성이 높습니다. 법적 책임과 비용 구조는 감독형 틀 안에서 통제하되, 안구 감시 강도를 대폭 낮춰 체감 사용성을 극대화하는 방향이 훨씬 합리적이니까요.
Tesla FSD v15 goes to wide release next month. Coincidentally Cybercab deliveries are also next month and Elon hinted about unsupervised FSD for Model Y fleet management in Austin next month too. So tonight was nothing special but next month is about to be crazy. Especially for those of us in Austin
@MichaelWaitze 한국에서는 2026.44.2가 발견되었습니다. 현재까지 4대이고 모두 기가상하이에서 생산된 차량입니다. 4대 중 3대는 FSD를 구입하지 않은 차량이고 한 대는 FSD를 구입했는데 아직 규제 때문에 배포받지 못한 차량입니다.
https://t.co/8uA7Ky9FQz
🔥 Tesla just teased the launch of FSD V15, possibly as soon as this weekend!
This is expected to be the version that enables unsupervised autonomous driving, starting in Q2 2027.
It could also be the version powering the first privately owned Cybercabs in December.