Tesla FSD v14.2.2.5 Is Rolling Out via Update 2025.45.10 — Here's Everything That Changed

By BASENOR Editorial · February 2026

Tesla's latest over-the-air push is live. Software update 2025.45.10 — carrying Full Self-Driving (Supervised) v14.2.2.5 — began hitting Hardware 4 vehicles in waves starting around February 14–16, 2026, with tracker Teslascope confirming a fresh deployment wave as recently as February 18.

The rollout is currently limited to HW4-equipped vehicles (AI4 hardware), which covers the refreshed Model Y (Juniper), Model 3 (Highland), and Cybertruck. Additional waves are expected to follow. If you haven't received the notification yet, you likely will within days.

📊 Key Figures

What's New in FSD v14.2.2.5

According to release notes verified by multiple sources, this update is unusually feature-rich for an incremental version bump. Here's what changed:

🧠 Upgraded Vision Encoder

Tesla upgraded the neural network vision encoder at the core of FSD. The practical result: better detection and handling of emergency vehicles, roadway obstacles, and human gestures (think crossing guards and construction workers). This is foundational — improvements here lift performance across every FSD scenario simultaneously.

🅿️ Arrival Options — Choose Where FSD Parks

One of the most requested quality-of-life features is finally here. FSD now surfaces an "Arrival Options" menu that lets you tell the car where to park when it reaches the destination. Choices include:

• Parking Lot
• Street Parking
• Driveway
• Parking Garage
• Curbside Drop-Off

This closes a long-standing gap where FSD would occasionally pick an inconvenient spot without user input.

🐌 SLOTH & 💀 MAD MAX Speed Profiles

Tesla has expanded its Speed Profile options with two new extremes. SLOTH keeps speeds conservative and prefers the right lane — ideal for nervous passengers or school zones. MAD MAX sits at the opposite end: higher cruising speeds, more frequent lane changes, and an aggressive but legal driving style. These join the existing profiles, giving owners more granular control over FSD's personality.

🚨 Emergency Vehicle Behavior

FSD now has dedicated logic to pull over and yield when it detects emergency vehicles with active lights and sirens — a legal requirement in most jurisdictions that was previously handled inconsistently. This is a meaningful safety upgrade.

🗺️ Vision-Integrated Navigation

Navigation and routing intelligence has been folded directly into the vision-based neural network. In practice, this means FSD can detect a blocked road in real time and reroute on the fly — without needing to fall back to a map-based detour process. According to release notes, handling of static and dynamic gates, road debris, unprotected turns, vehicle cut-ins, and school buses has also been improved.

🔧 Fault Recovery & Windshield Alert

FSD has improved its ability to manage and recover from system faults — reducing the jarring disengagements owners sometimes experience during degraded operation. A new alert has also been added to flag residue build-up on the interior windshield that could interfere with the forward camera. This is a practical reminder for owners who don't frequently clean the inside of their glass.

📊 Self-Driving Stats in UI

A new "Self-Driving stats" section has been added under Controls > Autopilot. Tesla hasn't detailed exactly what metrics are shown, but this gives data-hungry owners a dedicated dashboard for their FSD usage history.

🔭 The BASENOR Take

📰 Deep Dive

The naming convention here tells a story. A jump from 14.2 to 14.3 would signal a major architecture shift. Staying within 14.2.x means Tesla is refining rather than rebuilding — but the feature list for 14.2.2.5 reads more like a major release than a patch. The addition of Arrival Options and the dual SLOTH/MAD MAX speed profiles alone represent months of user feedback being addressed in a single drop.

The vision encoder upgrade is the change that deserves the most attention, even if it's the least visible. Neural network encoder improvements cascade across every downstream FSD behavior simultaneously. When Tesla improves how the car sees, everything from pedestrian detection to lane merging and emergency vehicle response benefits — without any of those behaviors needing to be individually retrained. This is how Tesla compounds its FSD improvements at a pace that's structurally different from software that relies on rules-based programming.

The emergency vehicle pull-over behavior is both a safety feature and, frankly, a compliance feature. In most U.S. states, failing to yield to an emergency vehicle is a moving violation. The fact that FSD now has dedicated logic for this — rather than relying on general obstacle avoidance — suggests Tesla is hardening the system against real-world legal requirements as it pushes toward unsupervised operation.

HW3 owners will note, once again, that this update is HW4-exclusive. The hardware gap between AI4 and HW3 continues to widen with each FSD release. If you're on an older hardware platform and are deep into Tesla's ecosystem, the upgrade path is worth revisiting — particularly as feature parity between generations becomes harder to maintain going forward.