Dexterous In-Hand Manipulation [control policy]
No estimate
No hardware requirements for this model
The weights for this model have not been published, so it cannot be downloaded or run on your own hardware at any size. It is reachable only through its provider, and no graphics card changes that.
On record
Full specification
Everything on record for this model. Most of it describes how it was trained rather than how it runs — useful context for judging how much work went into it, and how it compares with models built at a different scale.
Origin
Who built this model, where, and when it was published.
- Organisation
- OpenAI
- Organisation type
- Industry
- Country
- United States of America
- Published
- 1 August 2018
- Authors
- Marcin Andrychowicz, Bowen Baker, Maciek Chociej, Rafal Jozefowicz, Bob McGrew, Jakub Pachocki, Arthur Petron, Matthias Plappert, Glenn Powell, Alex Ray, Jonas Schneider, Szymon Sidor, Josh Tobin, Peter Welinder, Lilian Weng, Wojciech Zaremba
What it does
The problem areas the model was built for. A model can carry several of each.
- Domain
- Robotics
- Task
- Robotic manipulation
Size
How large the model is and how much data it was trained on. Parameters are the figure that decides whether it fits on a given graphics card.
- Parameters
- 3.2M
- Training data
- tokens
See Figure 2 - Deployed model (D) consists of a Control Policy Network (B) and Vision Network (C), trained separately. (Training of (B) also includes training of a Value Network that is, ignoring input-output size differences, identical in architecture to that of the control policy; the value network is not included during deployment). Control Policy Network: 3181588 (23552 + 3147776 + 10260) see Table 2, Figure 12, Table 10 given: - input: 22 out [Table 2, 15+3+4] - normalization: 22 out - FC-…
Training compute
The arithmetic performed to train the model, measured in floating-point operations. It is a measure of what the training run cost, not of how fast the finished model answers you.
- Training compute
- 2.2 × 10²⁰ FLOP
- How it was established
- Hardware
In experiments, runs were typically conducted for 50 hours of wall time, with up to 32 GPUs. Table 10 specifies V100 GPUs for training the policy. Assume V100 base variant, BF16, 0.3 utilization V100 tensor FP16 performance = 1.25e14 FLOP/s 32 GPUs * (0.3 * 1.25e14 FLOP/s/GPU) * (3600 s/1 hr) * 50 hr = 2.16e+20 FLOP
The training run
What it physically took to train: which chips, how many, for how long, and what that drew from the wall.
- Training hardware
- NVIDIA V100
- Chips used
- 8
- Wall-clock time
- 50 hours
- Power draw
- 5.0 kW
Availability
Whether you can obtain the model and run it on your own hardware, which is what decides if any of the graphics-card figures on this page apply.
- Weights
- Closed — provider access only
- Model access
- Unreleased
- Training code
- Unreleased
How it is classified
Labels the source dataset applies when tracking notable models, and how confident it is in the entry.
- Record confidence
- Confident
- Citations
- 2,156
Sources
Where this record came from and when it was last checked.
- Reference
- Learning Dexterous In-Hand Manipulation
- Last updated
- 25 May 2026
What the numbers mean
Where it came from
Dexterous In-Hand Manipulation [control policy] was published by OpenAI, in the country recorded as United States of America, during August 2018. The category the publisher falls under is industry.
It works in the domain of Robotics, and is recorded as performing the task of robotic manipulation.
This is a closed model: the trained values stayed with whoever produced them, and there is no local version to run.
Training and provenance
Training it took a computation budget of roughly 2.2 × 10²⁰ FLOP, on hardware recorded as NVIDIA V100. That figure measures what producing the model cost, and has no bearing on how fast it answers.
Answers
Dexterous In-Hand Manipulation [control policy] — common questions
Dexterous In-Hand Manipulation [control policy]— how much compute was used to train it?
Training consumed around 2.2 × 10²⁰ FLOP, on hardware recorded as NVIDIA V100. That measures what producing the model cost and says nothing about how quickly it answers once trained — inference speed comes from memory bandwidth, not from the training budget.
Dexterous In-Hand Manipulation [control policy]— what GPU do I need to run it?
None. This is a closed model — its weights were never published, so it cannot be downloaded or run on your own hardware at any price. It is reachable only through its provider.
Dexterous In-Hand Manipulation [control policy]— is it open source?
No. Its weights have not been published, so it exists only as a service controlled by its owner.
Dexterous In-Hand Manipulation [control policy]— how many parameters does it have?
It has a parameter count of 3.2M. See Figure 2 - Deployed model (D) consists of a Control Policy Network (B) and Vision Network (C), trained separately. (Training of (B) also includes training of a Value Network that is, ignoring input-output size differences, identical in architecture to that of the control policy; the value network is not included during deployment). Control Policy Network: 3181588 (23552 + 3147776 + 10260) see Table 2, Figure 12, Table 10 given: - input: 22 out [Table 2, 15+3+4] - normalization: 22 out - FC-RELU: 1024 out [Table 10, assume bias] - LSTM: 512 out [Table 10, assume bias] - FC (implied?): 20 out [Appendix C.1, Actions] parameters: - FC-RELU: 23552 [(22+1)*1024] - normalization: 0 [ignored] - LSTM: 3147776 [4*(1024+512+1)*512] - FC (Implied?): 10260 [(512+1)*20]. That figure is the total, and it is what decides how much memory the model needs — roughly half a gigabyte per billion at the compression most people use.
Dexterous In-Hand Manipulation [control policy]— who created it?
It was published by OpenAI, based in United States of America, an organisation categorised as industry.
Dexterous In-Hand Manipulation [control policy]— when was it released?
It was published in August 2018. Capability per parameter has improved considerably since, so a newer model of the same size is often the better use of the same hardware.
Dexterous In-Hand Manipulation [control policy]— what is it used for?
It works in the domain of Robotics, and is recorded as handling the task of robotic manipulation. A model can carry several of each, so these are the areas it was built for rather than a limit on what it will attempt.
The other direction
Looking at it from the other side?
This page starts from the model. If you already own a card and want to know everything it will run, start from the hardware instead.