My Self-Driving G-Wagon

comma's public segments, closed loop

comma's public two-camera segments in worldmodel-4B with comma's ideal car, a vehicle that follows each command exactly.

28 segments from 19 drives. Run on 8 of 19 releases. When the test changes, every release is run again.

Distance from the recorded path at 2 s

Lower is better. Bar: 95% interval from resampling whole drives; tick: mean. Rows are releases, newest first; a gap between two releases smaller than their intervals does not separate them.
1020cm
  1. f22e06e55 Nov 2026, RL residualnot run
  2. f22b8b3429 Oct 2026, RL residualnot run
  3. 6b5b705f17 Oct 2026, RL residual11.5 cm (5.5 to 21.9)
  4. N16 nav model15 Oct 2026, RL residual8.8 cm (4.7 to 16.0)
  5. MD v3 / Flip Flop14 Oct 2026, RL residual8.7 cm (5.6 to 13.5)
  6. Mountain Dew10 Oct 2026, RL residual9.6 cm (4.8 to 18.2)
  7. 870a48238 Oct 2026, RL residual9.8 cm (5.3 to 16.9)
  8. Cinque v318 Sep 2026, Big RL12.8 cm (5.7 to 24.9)
  9. Cinque v211 Sep 2026, Big RL11.4 cm (5.3 to 22.4)
  10. lebowski1 Jul 2026, lebowski14.8 cm (6.9 to 28.1)
  11. Not run on the 9 releases before lebowski, 19 Jan 2026 to 7 Nov 2023: WMI, Dark Souls 2, The Cool People's, North Nevada, Steam Powered, Filet o Fish, North Dakota, Certified Herbalist, Farmville.
The table: mean, interval and median

comma's public segments, closed loop

comma's public two-camera segments in worldmodel-4B with comma's ideal car, a vehicle that follows each command exactly. 28 segments from 19 drives for every model.

Mean over segments with its 95% interval (resampling whole drives), median beneath. Newest release first.
Modelfirst public · statusDistance from the recorded path at 2 scm
6b5b705f2026-10-17, Hugging Face checkpoint (unshipped)11.5 cm[5.5, 21.9]median 6.0
N16 nav model 96244b1f2026-10-15, Hugging Face checkpoint (unshipped)8.8 cm[4.7, 16.0]median 5.4
MD v3 / Flip Flop 1365126a2026-10-14, merged to openpilot master (2026-10-14), updated 2026-10-248.7 cm[5.6, 13.5]median 7.4
Mountain Dew 8d7366832026-10-10, shipped in openpilot v0.11.2 (master 2026-10-10)9.6 cm[4.8, 18.2]median 5.6
870a48232026-10-08, Hugging Face checkpoint (unshipped)9.8 cm[5.3, 16.9]median 5.6
Cinque v3 f78ed37d2026-09-18, merged to openpilot master (2026-09-18)12.8 cm[5.7, 24.9]median 6.7
Cinque v2 1a4211752026-09-11, merged to openpilot master (2026-09-11)11.4 cm[5.3, 22.4]median 6.3
lebowski2026-07-01, merged to openpilot master (2026-07-01)14.8 cm[6.9, 28.1]median 7.2
Conditions and limits

Measured. comma's worldmodel-4B generates each next frame at 5 Hz from 9 history frames, 5 recorded anchor frames and the car's pose; the model drives comma's ideal car (the requested curvature and acceleration happen at once) for 7.2 s from the recorded start. Distance is to the recording car's own path. One window per segment, nearest the middle, where the car holds at least 5 m/s for the window and the 4.8 s before it; segments with no such stretch are left out. Seeds 0 and 1, averaged per segment. 28 of the 45 listed segments have a score, from 19 drives and 17 cars.

Conditions and limits

Measured. comma's worldmodel-4B generates each next frame at 5 Hz from 9 history frames, 5 recorded anchor frames and the car's pose; the model drives comma's ideal car (the requested curvature and acceleration happen at once) for 7.2 s from the recorded start. Distance is to the recording car's own path. One window per segment, nearest the middle, where the car holds at least 5 m/s for the window and the 4.8 s before it; segments with no such stretch are left out. Seeds 0 and 1, averaged per segment. 28 of the 45 listed segments have a score, from 19 drives and 17 cars.

How this test has changed

  1. Closed loop on comma's public segments added: 28 segments in worldmodel-4B, comma's ideal car, seeds 0 and 1. commit f0ec83f