Preliminary observations of Dream-1 on a real bimanual robot

We ran dreamlabs/dream-1-beta on a bimanual YAM rig from 21 October to 5 November 2026: 98 recorded rollouts across marker handover, pick and place, bottles, cloth, cups and a loaded plate. We judged each run from the footage. For the 23 runs on the official Dream CLI, we also checked the control and inference logs.

In two runs, Dream-1 picked up a marker, passed it to the other gripper in the air and placed it in a box. GPT-6 Astra showed no transfer in the two partial recordings we have for this task. Across tasks, grasps were often lost or retried: the ten bottle runs used 41 gripper closures for 7 successes. These are early results from a small sample.

Full marker handover, 2× speed. Right arm picks, passes the marker to the left in the air, and the left places it in the black box. Run c8124b0c.

Loaded plate, 2× speed. One gripper pinches the rim and drags. The plate tilts, the glass falls over, and the second arm never helps. Run ee8d2983.

Marker handover
2/6
complete, with a transfer seen in 4 of 6. Astra: 0 of 2 recorded
Two bricks into box
5/31
both placed, and 8 more with one brick placed
Bottle upright
7/10
succeeded, only 4 of them within two grasps
Loaded plate
0/3
level lifts. All three runs used a one-sided rim pinch
Summary

What we ran and what we found

Dream-1 used the named arm in every bottle run and completed some two-arm transfers. Most failures came after contact: keeping a grasp, supporting a load, and stopping when the task was done.

TaskRunsAchievedPartialNot achievedNo verdictRead
Marker handover62301Transfer seen in 4 of 6, with placement in 2.
Two bricks into a box, one per arm3158126Results varied. Some grasps failed, and the arms sometimes shifted the box.
Stand a water bottle upright107030Correct arm in 10 of 10, but most runs need repeated attempts.
Fold the clothes once32001Both hems pinched, finish quality is coarse.
Stack the red cups50302First nest worked in all 3 full attempts. The third cup failed every time.
Cup into box, then stack171943Usually reaches the box. Rim grasps tip, hold or throw the cup.
Lift a loaded plate30030Rim pinch and drag. No stable support, no help from the second arm.
Single-arm brick into box80152Earliest runs on a different table setup. Mostly missed grasps.
Other early tasks1500105Uncapping, insertion, basket, towel, two bottles and others.
Total9817243720Visual judgments, not a benchmark score. Tasks vary widely in difficulty.

What we saw work

  • In-air marker handover with placement, in 2 of 6 runs.
  • Following the named arm in every single-arm bottle prompt.
  • Pinching both hems of a cloth and gathering it into a fold.
  • Inserting one cup into another.
  • Retrying after a failed attempt, without human reset.

Where runs failed

  • Grasps that slipped or were retried, and objects dropped or knocked away.
  • No two-sided support seen on wide, loaded objects.
  • Arms continuing after a task was done, or closing on nothing.
  • The same prompt succeeding in one run and failing in the next.

Possible next checks

  • Whether gripper targets close through the object or stop at its width.
  • A stop signal, or a success check on the wrist view.
  • Two-arm strategy on wide objects.
  • Round-trip latency: on the CLI, the robot waits for actions 24% of policy time.
Findings

Observations

Each item describes what the footage shows, with logs where we have them. Possible causes are marked as hypotheses.

1

Grasps were often lost or retried

This was the most common pattern. In pick and place, only 5 of 31 two-brick runs placed both bricks. Ten bottle runs produced 41 gripper closures for 7 successes. Only 4 runs succeeded within two closures, while others took 5, 6 and 10. Run 1636f966 closed 11 times, and then the bottle left the table. The same instability appears on cups, with a cup thrown off the table in two of three full attempts.

A possible contributor is grip force. Across 27,345 closed-gripper ticks, measured width sits a median of just 0.005 above the target. On position-controlled grippers, that small an error implies little squeeze. Our station also ran grip_force_cap: 0.65, which should be ruled out as a cause.

HypothesisIf that target is the raw model output, Dream-1 is predicting the width it sees instead of closing through the object.

2

Wide, loaded objects were pinched at the rim from one side

All three plate runs share a strategy: the right gripper pinches the rim, then drags and pivots the plate. The left gripper never closes on it. The plate tilts, and in ee8d2983 the glass on it falls over. The plate was pushed more than carried.

A fruit basket prompt that named both handles also shows no level two-arm lift in its two recorded runs. We saw two arms coordinate on a thin object, but not share a load.

CaveatThe plate prompt did not name both arms. A direct follow-up is the same plate with an explicit two-arm prompt.

3

Runs did not end on a policy stop, and some arms kept working after the task was done

None of the 23 official CLI runs ended on a policy signal: 19 were stopped by the operator, 3 by arm faults and 1 by a control loop error. In 8bbb3ddb, the arm returns to a bottle that is already standing, closes on nothing and stays there for 23 s. In 1636f966, the gripper keeps closing after the bottle has left the table. In a1324c12, the arm goes home with the gripper still shut and idles for three minutes.

NoteIn several runs the arms did go home after success, so some completion behavior may already be present.

4

Later steps failed more often than early ones

Cups: the first nest worked in 3 of 3 full attempts, and the third cup failed in all 3. In cab53793 the right arm places the marker on the plate directly, and only afterwards do both grippers close on air, far apart. In 32dd1777 the transfer works, then the left arm holds the marker over the box and never lets go.

HypothesisThe station config uses obs_window: 9 at obs_fps: 5, about 1.8 s of history. A window that short gives little direct evidence of which sub-goal is complete. We would value your view on how the model tracks progress.

5

Control often paused while waiting for actions

Median round trip was 962 ms, and 64% of requests exceeded the CLI's own 900 ms pacing threshold. Network accounts for about 73 ms of the median. Control ticks stalled for more than 200 ms during 24% of policy time, with stops as long as 13.9 s. Contact-rich motion that pauses while waiting for actions may contribute to dropped grasps. We have not isolated that effect.

NoteNot all of the instability is the model. Our left arm CAN faults ended 3 runs and blocked 8 launches. See Runtime.

Setup

Two YAM arms, three cameras

Runs span 21 October to 5 November 2026. Prompts are short and natural. The 75 runs before 3 November used our earlier integration, and the 23 after it used the official Dream CLI.

RobotBimanual I2RT YAM, 6-DoF per arm, linear parallel grippers
Cameras3 Intel RealSense: top, left wrist, right wrist. CLI runs recorded at 1280 × 720, 30 fps, with 640 × 360 model input
Control, CLI runs20 Hz policy ticks, 250 Hz motor commands, 48-action chunks, end effector capped at 0.6 m/s
Modeldreamlabs/dream-1-beta
Software23 runs on dreamcli robot run, CLI 1.0.5, client 1.0.3. 75 on our earlier integration
Scope98 YAM runs with usable video. Runs without saved video are not scored
Task 1 · 6 runs

Marker handover, and how it compares with Astra

TASK PROMPTPick up the marker with one arm, hand it to the other arm, then put it into the black box.

This was the most complete two-arm behavior in the set. The right arm picks up a thin, reflective marker, the two grippers meet in the air, the left takes it and the right lets go. The left then places it in the box. Both wrist cameras confirm the transfer.

Complete in under a minute, real time. Pick, transfer and place, with the marker in the box by 0:43. Run ead9ed03.
Transfers in both directions on the CLI, 2× speed. Left to right, then right to left, on 5 November. No black box was in view, a setup error on our side, so we cannot score placement. Run d0cabfde.

Dream-1 vs. GPT-6 Astra

Same YAM rig, recorded runs only. In both Astra recordings, the right arm picks up the object and the left does not take it in the saved footage.

PolicyRecorded runsTransfer seenPlaced after transfer
Dream-1642
GPT-6 Astra200
GPT-6 Astra, same prompt, 2× speed. The arms come together, with no transfer in the 50 s of saved video. Run 2311c01d.
GPT-6 Astra, transfer to plate, 2× speed. The right arm holds a highlighter while the left works around the plate, with no transfer in the 65 s of saved video. Run 55b94c09.

A small sample, so this is a direct observation on this task, not a general ranking. Scenes and objects differ between some runs. Astra's saved videos are also partial: 50 s of a 144 s run that was cancelled, and 65 s of a 251 s run, so a later transfer would not appear. Further Astra attempts have no saved video and are not counted. Dream-1's other runs: 32dd1777 transfers but never releases, 5cc32b20 ends on an error before transfer, and cab53793 skips the transfer.

A run without a transfer, real time. Prompt: right gripper, transfer to left, place on the green plate. The right arm places it directly, then both grippers close on air near 0:43. Run cab53793.

Task 2 · 31 runs

Two bricks into a box

TASK PROMPTPick up the two lego bricks, one with each arm, and put both of them into the black box, then call done.
5
runs with both bricks placed
8
runs with one brick placed
12
runs with no brick placed
6
runs ended early, no verdict
0 → 5
both-brick results, 22 Oct vs. 23 Oct, with scene changes between days

In the successful runs, the arms take turns. The same prompt also produced missed grasps, a second brick left behind and a box pushed out of position.

Both bricks placed, real time. Run 76588ec0.
Both bricks placed, 2× speed. Run a9411fd6.
One brick placed, box shifted, 2× speed. The arm disturbs the box while placing, and the second brick stays out. Run f74c027f.
Task 3 · 10 runs

Stand a water bottle upright

TASK PROMPTStand the water bottle upright using the right arm. (5 runs, and 5 more naming the left arm)
7/10
upright at policy stop
10/10
used the arm named in the prompt
41
gripper closures in total
1:13
median time to final release, in successful runs
2
runs where the bottle left the table
A one-grasp run, 2× speed. One grasp at the bottle's width, rotate, place, go home. Run 08bf7138.
A multi-grasp run, 4× speed. Grasp, rotate, set down, still not upright, try again. It gets there after six closures. Run 797b8ac3.

Most runs needed several grasps

Each row is one bottle run, sorted by number of closures. Each bar is one closure of the arm named in the prompt. Time runs from the start of the recording. Hover a bar for details.

Closed at object widthClosed almost fully, below 0.12: nothing substantial between the fingers

How the failures look

Bottle lost, gripper keeps closing, 2× speed. After repeated closures, the bottle leaves the table near 3:30. The next closures happen on air. Run 1636f966.
Swept off the table, 2× speed. After a 12 s stall and a server-side session reset at 0:48, the arm sweeps the bottle away with its gripper open and requests 0.83 m/s, above the 0.6 m/s cap. It then reaches past the table for about 16 s. The reset is close in time, but we have not shown it to be the cause. Run 9049349c.
Re-engaging a finished task, 2× speed. The bottle is already standing. The arm comes back and closes fully next to it. The bottle stays standing. Run 8bbb3ddb.
Task 4 · 3 runs

Fold the clothes once

TASK PROMPTFold the clothes once.

Both arms approach symmetrically, pinch the near hems at nearly the same moment and gather the shirt. Two of three runs end folded. The third was still taut when our left arm faulted, so we do not score it.

Compact fold, 2× speed. The arms release and go home on their own. Run 29b8119e.
Slower route to a coarse fold, 4× speed. About 15 s of no arm motion while pinching, then regrasps. The edge finish is uneven. Run a2be0488.
Task 5 · 22 runs

Cups: stacking, and placing into a box

TASK PROMPTStack the red cups.

The end states understate how far these runs got. In all three full attempts it inserts one cup into another, and in two runs it leaves a clean two-cup stack standing. The third cup was not placed in any of them.

A further 17 runs asked Dream-1 to put a cup in a black box and sometimes stack a second one on it. The cup reached the box in most fair attempts, but 4 were tipped or left held, and 4 others were tilted or thrown before arrival. Rim grasps on the light cups often failed, as in the stacking task. One run, cb95bd50, nested the cup cleanly.

Clean nest in the box, real time. Run cb95bd50.
Cup thrown off the rim grasp, real time. Run 953696a6.
Nest, carry, lose the third, 2× speed. The left wrist view shows the two rims clearly at the end. Run 3a8c040c.
Third cup thrown off the table, 2× speed. Run 1377aa64.
Repeated attempts, 4× speed. Regrasps a knocked-over cup and ends inverted. Run d3d24f72.
Task 6 · 3 runs

Lift a loaded plate

TASK PROMPTLift the green plate by 5 cm, move it 5 cm to the left, then 5 cm to the right, and place it back down.

The plate carries a glass and, in some runs, a small snack bar, so a stable lift would need support from two sides. Instead, the right gripper closes on the rim almost immediately and stays closed: 57 s, 91 s and 223 s across the three runs. The plate pivots and slides along the table. We never see a level lift, or a clean left then right motion.

Pinch and drag, 2× speed. The left arm stays parked. The right wrist view shows how little of the plate is held. Run 3c345a0a.
Returns home with the gripper closed, 4× speed. After early contact the arm goes home still closed. There are no further attempts in the remaining three minutes. Run a1324c12.

Fruit basket, both handles named in the prompt, 4× speed. The same limit on a different object: no level two-arm lift in the recording. Run e12bb8d2.

Failure modes

Where each problem sits

We separate model behavior from our test setup and from runtime infrastructure, so that only the first counts against Dream-1. The telemetry-based rows use CLI runs only.

Failure modeSourceRunsEvidence
Unstable grasp, repeated regrasp, dropModel, maybe grip forceMost commonAcross bottles, bricks and cups. Examples: 797b8ac3, 8bbb3ddb, 1636f966, 1377aa64, f74c027f
No termination, re-engages after successModel23 of 23 CLINo policy stop in any run. 8bbb3ddb shows re-engagement
No shared support of a wide loadModel5All plate runs, one-sided rim pinch. Both recorded basket runs
Disturbs or ejects another objectModel6+3a8c040c, 1377aa64, d3d24f72, 9049349c, 953696a6, f74c027f
Closes on nothing, no sign of noticingModel31636f966, 8bbb3ddb, cab53793
Sequence out of orderModel2cab53793 skips the transfer. 32dd1777 never releases
Motion above the speed cap, clampedModel, safety held59c9ab4cb, 9049349c, a2be0488, d3d24f72, 1377aa64 at 0.62 to 1.78 m/s against a 0.6 m/s cap. Excludes a 4.88 m/s reading at the 89dc8159 arm fault
Scene did not match promptOur setup1d0cabfde: no black box in view
Left arm CAN faultOur hardware389dc8159, a1324c12, a3aa2602. Also 8 failed launches, all from can_yam_eval motor 1
Session reset or connection errorRuntime, cause open49049349c, 6a1c5288 (session), a2be0488, d0cabfde (TLS EOF, recovered)
Runtime

Latency and stability across 1,429 CLI action chunks

962 ms
median round trip. p90 1,423 ms, p99 4,406 ms
318 ms
median serve_infer_ms
742 ms
median gateway_ms
73 ms
median network_ms. p95 284 ms
24%
of 3,355 s policy time in control gaps above 200 ms

The CLI selects sync or async pacing around a 900 ms median round trip. With a median of 962 ms, runs sit right on that boundary: 457 chunks ran in sync and 972 in async, often switching within a single run. So the motion profile can change mid-task. One cup run and the short marker run were fully sync. It may be worth testing Dream-1 with this switching held constant.

Field names are as logged in inference.jsonl. We did not separately measure our site's link capacity, so we treat network_ms as logged and do not claim to have fully isolated the network.

All runs

Every recording

Click a row to watch that run at 2× speed. Timestamps on this page are in recording time. Early runs have no telemetry columns.

RunVerdictWhat happensPolicy timeClosuresWaiting
Method and limits

How to read this

  • Verdicts are visual. We judged the state at the moment the policy stopped, not after homing or any human contact. We did not use stored outcome labels. Any error in judgment is ours, and we would welcome corrections.
  • Review depth. We reviewed every run on evenly spaced frames, with closer passes on key contact moments and final frames. We did not inspect every frame of every run, so brief events could be missed. For some long early runs, the saved video covers only part of the run, and we judged what was recorded.
  • Two integrations. Runs before 3 November used our earlier integration, not the official CLI. Telemetry findings come only from the CLI runs, and results across the two periods are not strictly comparable.
  • Small samples. 98 runs across varied tasks, tables and lighting, with short natural prompts. Treat counts as signals, not rates.
  • Operator stops. Almost every run ended by hand, so some attempts may have been cut short. Run length is partly our decision.
  • Video processing. The CLI recordings were dim. For viewing only, we applied one fixed tone curve and white balance per camera, with light denoise, and it never changes within a clip. Early recordings were already well exposed and are unprocessed. The model saw the original images.

Independent evaluation by Robotics Center, not produced or endorsed by Dream Labs. Dream Labs and Dream-1 are names of Dream Labs. Model tested: dreamlabs/dream-1-beta.