Manipulation · Basketball¶
Excluded from the benchmark
not among the hardest (robot_coding_bench 2026-10-04): passed in unlimited mode
Greyed and last in the task list, left out of the benchmark's numbers; its demo and agent runs stay (excluded: in state/tasks/humanoidbench.yml).
Agent runs
Each mode of this task runs once per run. Est. cost is tokens at list price (data/prices.yml), never a bill; see the HumanoidBench runs for every task of a run.
Codex CLI 0.157–0.159 + GPT-6 Luna, reasoning medium (OpenRouter)default run
| Trial | Started | Agent time | Model requests | Tokens in / out | Est. cost | Grade |
|---|---|---|---|---|---|---|
| U✗ 1h 00m | 09-28 18:18 | 1h 00m | — | 24.4M / 219k | $0.407 | deterministic 1, grader_error 0, n_actions 125, video_rendered 1 |
| L✗ 1h 00m | 10-01 15:06 | 1h 00m | — | 29.7M / 175k | $0.406 | deterministic 1, grader_error 0, live_success 0, missing_log 0, n_actions 51, replay_success 0, video_rendered 1 |
3 other run(s) of HumanoidBench
Codex CLI 0.159.2 + GPT-6.1 Sol, reasoning medium (OpenRouter)
| Trial | Started | Agent time | Model requests | Tokens in / out | Est. cost | Billed | Grade |
|---|---|---|---|---|---|---|---|
| U✓ 1h 00m | 10-01 03:46 | 1h 00m | 157 | 14.1M / 105k | $2.79 | $2.87 | deterministic 1, grader_error 0, n_actions 500, video_rendered 1 |
| L✗ 58m | 09-30 23:27 | 58m | 160 | 14.6M / 81k | $2.58 | $2.67 | deterministic 1, grader_error 0, live_success 0, missing_log 0, n_actions 177, replay_success 0, video_rendered 1 |
Claude Code 2.1.283 + Claude Opus 5.5, reasoning medium (OpenRouter)
Not in this run: not in this run (the two dearer models ran five tasks, chosen to span locomotion and manipulation, easy to hard)
Codex CLI 0.157.0 + GPT-6 Sol, reasoning medium (OpenRouter)closed
Not in this run: not in this run (the two dearer models ran five tasks, chosen to span locomotion and manipulation, easy to hard)
Task instruction (upstream)
A ball is thrown at the robot. Catch it, then throw it through the hoop.
What HumanoidBench states about this task
| Success criteria | 1. the summed per-step reward over one episode reaches 1200, HumanoidBench's own success bar (a total of rewards, not a number of steps; an episode is at most 500 control steps) 2. unlimited: both fresh-process replays of the handed-in trajectory reach it and end in the same state 3. limited: the run passes the moment a live episode reaches it (the recorded episode replays to the same state); otherwise the last episode is graded |
| Env Id | h1-basketball-v0 |
| Robot | Unitree H1 (19 actuators) |
| Category | Manipulation |
| Capability Class | C5 · whole-body power and momentum |
| Role | scored |
| Scoring | Scoring runs in two stages and switches by itself the first time the robot contacts the ball. Before that it rewards standing steadily and getting both hands to the ball; after it, the dominant term is how close the ball is to the hoop. Sinking it pays a one-off bonus that covers most of the bar but not all of it, and ends the episode immediately, so the rest has to be earned per step before the ball goes in. That bonus is not optional: no amount of good positioning reaches the bar without it. |
| Ends Early | The episode also ends if the ball drops low or the robot falls. |
| Zero Action Return | 10.77 |
| Action Dim | 19 |
| Control Rate Hz | 50 |
| Limited Mode | head cameras (RGB 256×256), joint angles and velocities; a pelvis IMU and a camera fixed in the room when the run turns them on. Not where the robot is in the room, not where the ball or the hoop is, not the reward |
| Agent Budget | 3600 s of wall clock per mode |
From https://github.com/carlosferrazza/humanoid-bench @ cb11890, as defined in our task definitions @ 7d6a7a44.
Tags¶
Why this task is interesting¶
Catch a ball thrown at the robot at about 7.5 m/s, then throw it through the hoop. The basket pays a bonus, but the rest of the bar has to be earned step by step before it.
Capability notes¶
Not yet written.
Oracle demo review¶
No demo.
Discussion¶
Passes in unlimited mode by holding the ball to earn the per-step reward and scoring on the last step; limited mode catches and throws, but always short. (@williamzhangNU)