Tesla FSD 14.3.2 Completes NYC to LA Cannonball Run with Zero Interventions | State of Self-Driving Cars 2026

The Cannonball Run is the most culturally loaded benchmark in American motoring, a coast-to-coast test that has defined the limits of driver, machine, and road since Brock Yates organized the first illegal run in 1971. That a consumer software update on a production Model 3 could complete the full route without requiring a human hand is a different kind of record than anything the original runs envisioned.

The team departed the Red Ball Garage on East 31st Street in Manhattan — the traditional start — and arrived at the Portofino Inn in Redondo Beach, the traditional finish, 49 hours and 55 minutes later. Across that time, FSD v14.3.2 managed all highway driving, all urban navigation in and out of New York and Los Angeles, and autonomous parking maneuvers at each Supercharger stop. At no point did the human occupants intervene with the wheel or pedals. Dan Burkland confirmed on X: "We leveraged FSD v14.3.2 for the ENTIRE trip, including parking at every charging stop, all without a SINGLE disengagement or intervention."

Tesla's official account reposted the run with the description: "Tesla on FSD Supervised drives itself from NYC to LA with zero interventions." The phrasing — FSD Supervised — is deliberate. Tesla is careful to distinguish the supervised Level 2 classification from unsupervised autonomy. Burkland's team operated within that classification, with attentive occupants, while exercising the full capability of the system. For full Elon Musk and Tesla news coverage see the ObjectWire hub. Analysis by Jack Sterling, Autonomous Vehicles Correspondent.

The zero-intervention FSD Cannonball category is barely four months old. Alex Roy, a veteran of multiple Cannonball records and one of the most credible voices in the autonomous vehicle enthusiast community, set the first zero-intervention mark on January 22, 2026. His team drove a 2024 Model S from Los Angeles to New York in the reverse direction, covering 3,081 miles in 58 hours and 22 minutes on FSD version 14.2.2.3.

The January run was notable for its conditions: Roy's team encountered snow, ice, and slush for meaningful stretches of the route, conditions that historically expose the limits of camera-only autonomy systems. FSD 14.2.2.3 completed it without a single disengagement, a finding Roy reported publicly. He was also clear about the classification: supervised Level 2 operation, human-attentive throughout. The run was the first confirmed proof of full-route FSD completion at Cannonball scale.

Roy responded to Saturday's record being broken by noting that FSD records are following the "pioneer-to-optimization" pattern of classic Cannonball history, in which the first completion establishes possibility and subsequent teams optimize time through better route planning, charging strategy, and software iteration. He remains the pioneer. Burkland and team are the first optimizers. The record will likely be broken again before FSD 15 ships.

Tesla released FSD version 14.3.2 in late April 2026. The update introduced three substantive changes relevant to a coast-to-coast run: upgraded reinforcement learning from expanded fleet edge-case data, an improved neural vision encoder optimized for low-visibility scenarios including night driving and highway glare, and approximately 20 percent faster reaction times in obstacle response calculations. Tesla also unified the AI model architecture across consumer FSD, Smart Summon, and the robotaxi deployment pipeline, meaning the same model weights serve all three products rather than maintaining separate training branches.

The reaction time improvement is the likely direct contributor to the record margin. A 20 percent faster response across 2,833 miles of variable traffic, merging, and interstate lane changes compounds meaningfully into total run time, independent of route planning or charging stop efficiency. The neural vision encoder upgrade addressed a known limitation of the v14.2.x series in sustained low-light highway driving, which the January route through winter northeast conditions had partially tested but which a May cross-country run tests differently through desert night segments.

The unified model architecture is the longer-horizon signal. Tesla is converging its consumer FSD product with its robotaxi platform model weights, meaning every improvement to one directly improves the other. When the robotaxi fleet accumulates edge-case data at scale, those training signals flow back into consumer FSD. The Cannonball run was performed on the same model weights Tesla intends to deploy in its unsupervised robotaxi service.

The Tesla Cannonball record is a highly visible consumer demonstration, but the autonomous driving landscape in mid-2026 is more stratified than any single benchmark suggests. The key distinction that explains nearly every difference between platforms is SAE level: Level 2 (human supervises, system controls), Level 3 (system supervises in defined conditions, human available), and Level 4 (fully driverless within operational design domain, no human required).

Waymo One is the most commercially mature Level 4 deployment in North America. Its Phoenix, San Francisco, and Los Angeles robotaxi services run 24/7 driverless using lidar, radar, and camera sensor fusion. Waymo's operational design domain is geofenced, meaning it operates in defined mapped areas and does not yet handle arbitrary interstate navigation at Cannonball scale. Within its zones, Waymo is further along the commercial autonomy curve than any other platform. The company completed over 1 million paid driverless trips as of Q1 2026.

Zoox, Amazon's autonomous vehicle subsidiary, is developing a purpose-built bidirectional robotaxi with no steering wheel or pedals. Testing in Las Vegas and San Francisco is ongoing, with commercial service targeted for late 2026. Aurora launched the first commercial driverless long-haul trucking operation in April 2024, running freight corridors between Dallas and Houston. The commercial trucking sector is further advanced on Level 4 deployment than the passenger robotaxi sector because highway-only operational design domains are more tractable than mixed urban environments.

GM Cruise paused all driverless operations in October 2023 following a safety incident involving a pedestrian and has been in gradual re-permitting through 2025 and 2026. As of May 2026, Cruise is not at commercial scale and has no announced date for full driverless service resumption. The incident cost the platform roughly two years of competitive progress.

Tesla's position is unusual: it operates the largest deployed semi-autonomous fleet globally (millions of vehicles on FSD) while not yet holding Level 4 certification or operating a commercial robotaxi at scale. The Cannonball run demonstrates what Level 2 supervised capability looks like at its ceiling, while the June 30 European Commission regulatory review will determine whether FSD can progress toward Level 3 homologation in Europe. For all autonomous vehicle and tech platform analysis see the ObjectWire tech hub.

The immediate regulatory event for Tesla's FSD program is the European Commission Technical Committee on Motor Vehicles scheduled for June 30, 2026. The TCMV review covers Tesla's FSD Level 3 homologation filings, which would allow limited conditional autonomy in Europe without a mandatory attentive driver. Level 3 approval in Europe would be a significant legal and commercial step, enabling Tesla to offer hands-free operation under defined conditions without requiring the SAE Level 2 supervised posture its current service requires globally.

In the United States, the NHTSA advanced notice of proposed rulemaking on automated driving systems, published in Q4 2025, is working through public comment. No federal Level 3 or Level 4 homologation framework has been finalized. California's DMV autonomous vehicle regulations remain the most developed state framework, with Waymo and Cruise operating under California permits. Tesla does not hold a California AV permit and does not operate FSD under the AV permit framework, as FSD's Level 2 classification keeps it under standard vehicle safety rules.

The commercial trajectory for the broader industry by late 2026: Waymo One expected to expand geofenced coverage to Austin and additional markets. Zoox commercial launch in Las Vegas targeted. Aurora expanding freight corridors. Tesla robotaxi service in Austin and San Francisco targeted for late 2025, now tracking toward late 2026, using the same FSD model weights that completed Saturday's Cannonball run. The gap between consumer FSD capability and commercial driverless deployment is narrowing, but the legal and regulatory framework has not kept pace with the technical progress.

Is Tesla FSD 14.3.2 fully autonomous after this Cannonball run?

No. Tesla FSD v14.3.2 remains SAE Level 2, which requires an attentive human driver at all times. The Cannonball run demonstrates the system's capability to complete 2,833 miles without requiring any intervention, but it was performed with human occupants legally responsible for the vehicle. Fully autonomous, unsupervised operation at this scale would require Level 4 certification, which Tesla does not currently hold. The run is an extraordinary Level 2 demonstration, not evidence of Level 4 capability.

How does Tesla FSD compare to Waymo in mid-2026?

They solve different problems. Waymo operates fully driverless Level 4 robotaxi service in defined geofenced urban zones, with no human required in the vehicle. Tesla FSD operates at Level 2 supervision across arbitrary road conditions but requires attentive human oversight. Waymo cannot drive the Cannonball Run outside its operational design domain. Tesla FSD can navigate arbitrary routes but is not legally driverless. The comparison is less head-to-head and more parallel tracks toward different commercial deployment models.

What makes FSD v14.3.2 faster than v14.2.2.3?

Three improvements: upgraded reinforcement learning from fleet edge-case data accumulated since January, an improved neural vision encoder for low-visibility and challenging light conditions, and approximately 20 percent faster obstacle reaction times in the inference pipeline. Tesla also unified the model architecture across consumer FSD and robotaxi deployment targets in v14.3.2, meaning the same model weights serve both products and improvements to one improve the other.

When will Tesla launch a fully driverless robotaxi service?

Tesla has targeted Austin and San Francisco for robotaxi launches using the same FSD model weights as v14.3.2. Initial timelines pointed to late 2025, with current tracking suggesting late 2026 for a limited commercial rollout. The service would operate under supervised conditions initially before regulatory approval for unsupervised driverless operation. The European Commission TCMV review on June 30, 2026 is the next major regulatory checkpoint for FSD Level 3 homologation outside the United States.