Locomotion · Reach¶
Excluded from the benchmark
not among the hardest (robot_coding_bench 2026-10-04): passed in both modes
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✓ 59m | 09-28 18:19 | 59m | — | 26.5M / 151k | $0.362 | deterministic 1, grader_error 0, n_actions 1000, video_rendered 1 |
| L✗ 20m | 10-01 15:32 | 20m | — | 5.5M / 71k | $0.107 | deterministic 1, grader_error 0, live_success 0, missing_log 0, n_actions 1000, 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✓ 59m | 10-01 03:47 | 59m | 124 | 10.5M / 107k | $2.46 | $2.55 | deterministic 1, grader_error 0, n_actions 1000, video_rendered 1 |
| L✓ 16m | 09-30 22:25 | 16m | 56 | 4.6M / 32k | $1.06 | $1.13 | deterministic 1, grader_error 0, live_success 1, missing_log 0, n_actions 738, replay_success 1, 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)
Bring the robot's left hand to the target point and keep it there for as much of the episode as possible.
What HumanoidBench states about this task
| Success criteria | 1. the summed per-step reward over one episode reaches 12000, HumanoidBench's own success bar (a total of rewards, not a number of steps; an episode is at most 1000 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-reach-v0 |
| Robot | Unitree H1 (19 actuators) |
| Category | Locomotion |
| Capability Class | C8 · mobile manipulation |
| Role | scored |
| Scoring | There are two tiers, both paid per step rather than once on arrival: a moderate amount while the hand is within about a metre of the target, and a considerably larger amount on top while it is within a few centimetres. Staying upright is paid every step as well, and fast joint motion costs a little. Reaching the loose tier and holding it for the whole episode falls short of the bar — a meaningful fraction of the episode has to be spent inside the tight one. |
| Zero Action Return | 275.22 |
| 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 its hand is, not the target's coordinates, 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¶
The left hand must reach a target and stay within a few centimetres of it long enough, while the robot stays up for all 1000 steps; even holding the starting pose falls after about 100 steps.
Capability notes¶
Not yet written.
Oracle demo review¶
No demo.
Discussion¶
Passes in both modes, limited only just (12008 against 12000): a centre-of-mass crouch from spec() and a stereo-camera target let the arm reach without falling. (@williamzhangNU)