Tesla FSD v14.3.13 Rollout Widens With 2026.38.5.11 Update

A bug-fix build becomes the main release

Tesla's newest Full Self-Driving build is reaching cars quickly. Software version 2026.38.5.11 ships FSD (Supervised) v14.3.13 to newer Hardware 4 (AI4) vehicles and FSD (Supervised) v14.3 Lite to older Hardware 3 cars, and Tesla first pushed it on Saturday, October 24, before widening the release to many more vehicles across both hardware generations.1 Fleet tracker Teslascope counted about 6,380 more vehicles receiving the update within a five-minute window on October 26. In that wave, 64% of recipients were eligible AI4 cars getting the full v14.3.13 build.14

The update brings almost nothing new on paper. Not a Tesla App reported that the release notes and FSD version match the earlier 2026.38.5.10 build, which makes 2026.38.5.11 a refreshed binary rather than a new feature drop.9 The Tesla Newswire account went further and noted that even the v14.3.12 notes were carried over unchanged, calling it a major bug-fix update.5 Tracking sites describe this as the third or fourth wave of the v14.3.13 push, after 2026.38.5.10 began going out around October 20.45

The build matters because Tesla is using it to bring both hardware tiers onto one software branch. That branch carries the 2026.38 "Halloween" features and puts a current FSD build on HW4 and HW3 cars in the same package.514 Expect this to be the build most North American FSD owners run into November.

How far the rollout has reached

The trackers measure different fleets in different ways, so their adoption figures vary widely. Tesla Oracle cited TeslaFi and Tessie data showing the update on about 10% of tracked vehicles.1 TeslaMagz showed an 11% rollout, with installs rising from 5 on October 24 to 84 the next day and 280 on October 26.3 Teslascope's dashboard put 2026.38.5.11 at 16.3% of its North American fleet, making it one of the most widely distributed active releases.7 Tesla-info.com reported a much lower 3.22% of cars.6

None of these numbers should be read as the true share of Tesla's fleet. Each tracker has its own sample of owners and its own way of counting. They do agree on the trend: installs went from a trickle among early-access cars to a steep climb within about 48 hours, which is how Tesla's FSD releases usually move from cautious to confident. Teslascope also listed the release as North America only, and tesla-info saw it in the US and Canada.67

What v14.3.13 does on HW4

The AI4 release notes are long, even if they are inherited from earlier builds. Tesla says it rebuilt the AI compiler and runtime from scratch on MLIR. It credits that change with 20% faster reaction times and faster model iteration.12 The notes also describe an upgraded reinforcement learning (RL) training stage and a better vision encoder, aimed at rare and low-visibility scenes, 3D geometry and reading traffic signs.6

Most of the behaviour changes target the kinds of cases that tend to cause disengagements:

• Rare road users: better responses to emergency vehicles, school buses and drivers who violate right-of-way.2
• Animals and intersections: better handling of small animals and of complex traffic lights, including stopping for yellow lights. Tesla attributes both to RL training on hard examples taken from its fleet.1
• Lane behaviour: fewer cases of drifting to one side of the lane for no reason and fewer minor tailgating habits.1
• Degraded sensors: better handling of temporary system degradation, with the car staying in control and recovering without a driver takeover.6
• Driver monitoring: a more sensitive camera system with improved eye-gaze tracking and better handling of glasses.8

One line in the notes stands out as strategy: Tesla says it has unified the model used by Actually Smart Summon, FSD and Robotaxi.6 The company now treats its consumer driver-assist and its commercial ride-hailing work as one software stack. Owner-facing additions point the same way. These include parking spots shown on the map with a (P) icon, a counter for distance driven without an intervention, a prompt to pick a reason after taking over, and a higher 8 mph top speed for Actually Smart Summon.18 The intervention prompts and streak counters work as a data-collection tool presented as a feature. Tesla is asking owners to label their own takeovers, which can feed the training process the release notes keep describing.

AI4 cars also get Automatic Collision Evasion. When the driver is in manual control, it can switch on FSD to try to avoid a crash, for example when a frontal collision is imminent and braking alone may not prevent it.13 Tesla Oracle presents this as new in 2026.38.5.11.12 However, Basenor's coverage shows the feature was already part of the v14.3.12 Halloween release earlier in October.14 For most HW4 owners, this update matures the feature rather than introducing it.

Tesla's list of upcoming improvements includes extending reasoning to all driving behaviours, not only destination handling, and adding pothole avoidance.68

The HW3 story: real progress, clear limits

For HW3 owners, this release continues a slow recovery. Tesla began offering FSD v14 Lite to HW3 Model 3 and Model Y in June, using distillation to train the older computer on the behaviour of the much larger AI4 model.17 The v14.3 Lite notes describe the same method, saying it brings over HW4 gains such as RL improvements. They also list smoother steering, fewer false slowdowns, and new parking, unparking and reversing abilities.11

The rollout has been uneven across HW3 cars. Older Intel-based Model S and X cars got their first v14 Lite build only on October 14, in update 2026.38.5.1.1619 On those cars, Teslascope's early feedback was described as a big step up from V12, with better speed control and fewer steering-wheel nag alerts. The same feedback noted that parking, arrival options, reversing and Actually Smart Summon were missing.14 TeslaMagz's feature breakdown shows Arrival Options and UI improvements only for some HW3 cars.3 Tesla Oracle notes that Intel-based cars mainly gain Speed Profiles.11

The reports disagree on two points. The first is AMD Ryzen-based HW3 Model S and X. Autoevolution wrote in mid-October that those cars were still waiting16, while Not a Tesla App reported they had started getting v14 Lite in September.19 The second is whether 2026.38.5.11 carries a new HW3 build at all. Teslascope's dashboard tags the release only with AI4 v14.3.137. Meanwhile, Tesla Oracle reports legacy S/X owners receiving v14.3 Lite through it12, and tesla-info has seen the Lite build on HW3 Intel cars.6 The most likely explanation is that HW3 cars get the shared package with an unchanged Lite model. Basenor describes it that way, saying HW3 owners keep getting v14.3 Lite "unchanged from prior waves."4 On that reading, HW3 owners are getting the latest branch, not a better HW3 driver.

There are practical concerns too. Autoevolution reports that v14 Lite puts heavy load on HW3 computers and can cause overheating and the "red hands" error. It suggests cleaning the cooling radiators before relying on FSD in older cars.16

Why it matters

The larger point is that v14 Lite looks like the last major FSD software line for HW3. Not a Tesla App reports Tesla has confirmed HW3 will not support unsupervised FSD or the upcoming v15 branch. Tesla has instead promised hardware upgrades for owners who paid for FSD.19 Basenor adds that Elon Musk acknowledged in April that HW3 cannot run FSD at an unsupervised level. All of these builds remain SAE Level 2 driver assistance that needs constant supervision.18

That produces a two-tier result. HW4 cars get a fast-moving stack shared with Robotaxi, with compiler rewrites, new safety features and a published roadmap. HW3 cars get distilled versions that are clearly better than V12, but with features that depend on which processor is in the car. They are fundamentally capped. The 2026.38.5.11 rollout shows Tesla can keep both groups on one release schedule. It does not close the capability gap, and Tesla's own roadmap suggests the gap will grow.

The rollout pace is also worth noting. Tesla went from early access to a wide release in about two days on a build whose notes had not changed. That suggests the company is confident in v14.3.13's stability. For an update meant to fix bugs, that confidence matters more than any single new feature.