Figure AI Edges Out Tesla Optimus With a 4.0 m/s Sprint Record

On December 3, Tesla posted a video of its Optimus robot running on X, captioned simply: "Just hit a new PR (Personal Record) in the lab."

The footage reveals that Optimus has achieved a highly naturalistic running gait. While Tesla has not officially confirmed the exact speed in the clip, the progress is stark when contrasted with the tentative, slow-paced walking demonstrations of 2023. In just over two years, Optimus has evolved from a stumbling prototype to a machine capable of a fluid run.

source: Tesla

Not to be outdone, Figure AI founder Brett Adcock responded yesterday by posting footage of the Figure 03 performing a running test—specifically, a shuttle run. Adcock noted that the Figure 03 controller is now driven by onboard neural networks built by the Helix team, adding that "the gap between human and robot performance is narrowing—rapidly and irreversibly."

Launched in October this year, the Figure 03 is positioned as a general-purpose humanoid for the home, promising to handle complex chores like folding laundry, washing dishes, and decluttering.

source: Figure

As the leading heavyweights in the American robotics sector, Tesla and Figure AI may diverge in their hardware and software philosophies, but both are accelerating toward a singular goal: dominating the dual holy grails of industrial automation and household service.

In terms of hardware, the Figure 03 stands 5'6" (approx. 1.68m) and weighs 132 lbs (60kg). It boasts a runtime of up to 5 hours on a full charge, supported by a dedicated wireless charging platform.

With Tesla Optimus clocking a record 3.8 m/s and Figure 03 achieving a top speed of 4.0 m/s in the shuttle run, the seemingly negligible 0.2 m/s gap is far more than a numbers game. It signals that humanoid robotics has officially crossed the "toddling" red line, entering the "deep water" zone where robotic dynamic mobility rivals that of humans.

For years, humanoid walking demonstrations were essentially exercises in "risk mitigation"—careful steps to avoid falling. This time, both companies demonstrated true running. In physics, the definition of "running" is strict: there must be an aerial phase where both feet leave the ground simultaneously. This means the robot is no longer "walking," but managing a continuous cycle of "controlled falling" and "ballistic launch."

Tesla’s "Blind Run": Extreme Balance in the Dark The Tesla team disclosed a hardcore technical detail: Optimus’s run was achieved in a "blind" state.

• The Technical Principle: The robot did not rely on vision to build a map or judge the terrain. Instead, it completely decoupled visual input, relying solely on proprioception—sensing joint angles via encoders and acceleration via the IMU (Inertial Measurement Unit).
• The 2-3 Millisecond Margin: At 3.8 m/s, any misjudgment of the center of gravity could lead to a catastrophic fall. Optimus’s neural network latency has been compressed to an astonishing 2-3 milliseconds. This is akin to a human sprinting on a treadmill with their eyes closed, relying purely on the cerebellum’s muscle memory and rapid-fire reflexes to maintain balance. Tesla is proving a point: if the model is robust enough, the laws of physics can be "internalized."

Figure’s "Shuttle Run": The Battle of Dynamics and Friction Figure 03 demonstrated strength in a different dimension. It wasn't just fast (4 m/s); it executed a high-difficulty shuttle run.

• Lateral Control: Linear acceleration only requires overcoming air resistance and inertia. A shuttle run demands emergency braking, pivoting, and re-acceleration. This requires the robot’s ankle joints to possess extreme lateral stiffness to withstand immense centrifugal force, while tire-grade rubber soles provide sufficient friction. The posture control displayed by Figure 03 during the turn reveals a high degree of synergy across its whole-body algorithms.

If the leg movements are the "external martial arts," the neural networks controlling them are the "internal energy." This brings us to the fiercest point of contention between the two firms: the battle of AI architecture.

Tesla: The Aesthetic of Data Brute Force

Tesla pursues a pur