A YouTuber just put Tesla FSD v14 Lite and other Full Self-Driving configurations to the test across three different hardware and software combinations. The results show just how much FSD has improved, and how much hardware still matters.
How the Test Was Set Up
The test was conducted by TechGeek Tesla on YouTube.
The location: a closed commercial park road with minimal weekend traffic — controlled enough to get clean data.
The method: a custom mannequin attached to a pulley system was pulled into the lane to simulate a pedestrian suddenly stepping into the road.
Reaction time was measured using dedicated telemetry software that tracked the vehicle's longitudinal acceleration. The clock started the moment the mannequin began moving, and stopped when the car's acceleration curve showed a braking or steering response.
Three Configurations Tested
Configuration |
Software Version |
Avg. Reaction Time |
|---|---|---|
HW3 |
v12.6.4 (baseline) |
1.065 seconds |
HW3 |
v14 Lite |
0.675 seconds |
HW4 |
v14.3.4 |
0.450 seconds |
The headline number: v14 Lite on HW3 is 37% faster than v12.6.4 on the same hardware. And HW4 with v14.3.4 is another 33% faster than v14 Lite.
For more background on how Tesla brought the V14 architecture to older hardware, see our FSD v14 Lite breakdown.
How They Controlled the Variables
Consistency was key to making this test meaningful.
The mannequin was pulled at the same timing point for every run
Each vehicle drove through at the same speed
A spotter stood roadside and signaled the puller the moment the car reached a specific position
Multiple runs were completed per vehicle to establish a reliable average baseline
Reaction time was calculated from the exact moment the mannequin moved to the first detectable change in the vehicle's longitudinal acceleration curve.
Some data was lost on test day due to phone overheating in high temperatures and dashcam footage splicing issues. But enough valid runs were captured to draw solid conclusions.
What the Results Actually Mean
The performance gap between versions is significant — and it's not just about software.
Elon Musk has previously stated that HW3's memory bandwidth is only 15% of HW4's. That makes v14 Lite's improvement on HW3 even more impressive.
The difference between the two hardware generations is also explored in our HW3 vs. HW4 FSD comparison, which looks at how the hardware gap affects the overall FSD experience.
So what exactly is FSD v14 Lite?
It's a trimmed-down version of Tesla's main v14 architecture, specifically optimized for the memory-constrained HW3 computer. The goal is to bring HW3 vehicles as close as possible to the current HW4 feature set — including:
Launching FSD from a parked state
Parking options upon arrival at a destination
Improved active and passive responses across navigation, lane changes, merges, pedestrian interactions, traffic lights, and cut-in scenarios
Tesla has been upfront that v14 Lite may be noticeably slower than the full v14 in some areas, given the hardware gap. The test results confirm this — v14 Lite sits between v12.6.4 and full v14 in reaction speed, which is exactly what you'd expect from a transitional solution while Tesla figures out a hardware upgrade path for older vehicles.
Hardware Still Matters — And Tesla Is Pushing Further
This test highlights something important: software and hardware need to evolve together.
HW4's superior compute power gives it the fastest reaction times. But v14 Lite proves that smart software optimization can still unlock meaningful gains on older hardware.
Tesla isn't stopping here. The company has announced plans to double the memory capacity on new HW4 units. And its upcoming AI5 chip will be significantly more powerful still.
More compute means the ability to run denser inference models — the technology behind those split-second driving decisions measured in this test.
As Tesla continues pushing both software updates and hardware generations, the capability gap between older and newer vehicles will be worth watching closely.