Drake
Drake holds place 7 of 13 in our robotics simulation software ranking. The catalogue describes it as: model-based design and control toolbox for C++ and Python, from MIT.
Vendors can pay for more visibility on this site; the link to Drake is marked like every other vendor link. How that works.
Visit Drake See it in the robotics simulation software ranking
Drake at a glance
| Headquarters | Not stated by the vendor |
|---|---|
| Free plan | No free plan recorded |
| Deployment | Unclear from the vendor pages |
| Pricing model | BSD 3-Clause licensed; plans and prices do not appear in the documentation read |
| Best for | Controls engineers who want analytical gradients and planning tools in Python |
| API | An API is mentioned by the vendor |
What is Drake?
Drake is a C++ and Python toolbox for analysing robot dynamics and building control systems, organised around Systems, Solvers and Multibody APIs. These model dynamical systems, solve mathematical programs and compute multibody kinematics and dynamics. The site says the software exposes the structure of its equations, including sparsity, analytical gradients, polynomial structure and uncertainty quantification, for planning, control and analysis algorithms.
It simulates robot dynamics including friction, contact and aerodynamics. The credits page says the project was started by members of the Robot Locomotion Group at MIT and the Robotics Division at Toyota Research Institute. Source is hosted on GitHub under the RobotLocomotion organisation, and the BSD 3-Clause licence text states that all components are licensed under it, with some portions under other permissive licences. The pages we read show no pricing and give no country of establishment.
Drake key features
- Systems, Solvers and Multibody. These three APIs split the work into modelling dynamical systems, solving mathematical programs and computing multibody kinematics and dynamics. Users therefore build models and controllers in one consistent set of tools.
- Visible equation structure. The site says Drake exposes sparsity, analytical gradients, polynomial structure and uncertainty quantification. This matters for people writing planning and control algorithms, though the pages do not give worked examples.
- Contact and friction simulation. The home page says Drake simulates robot dynamics including friction, contact and aerodynamics. Details of the contact models are left to the documentation beyond the pages we read.
- Python interface for prototyping. A Python interface supports rapid prototyping, alongside open-source implementations of state-of-the-art algorithms. Tutorials are Jupyter notebooks that can be run locally after installing the pip package.
- Several install routes. Pip, APT, Binary Download and Build From Source are listed. The pip route uses a virtual environment with pip install drake[mosek], where the mosek extra may be left out without a MOSEK licence.
- Nightly wheels. Wheel packages for Linux and macOS are published nightly, and the pip page gives commands to install a dated nightly wheel from the Drake package index.
- Tested configurations. The installation page lists tested setups including Ubuntu 24.04 LTS and 26.04 LTS on x86-64 and arm64, macOS Tahoe (26) and Golden Gate (27) on arm64, and Amazon Linux 2023.
Who Drake suits
- A controls researcher designing optimisation-based controllers. Access to gradients and mathematical program solvers lets the researcher formulate trajectory optimisation problems directly against a multibody model of the robot.
- A graduate student working through tutorials. Python notebooks in Jupyter run locally after a pip install, so a student can follow examples on a laptop without building the C++ source.
- A robotics team simulating contact-rich tasks. Friction and contact simulation help a team test manipulation behaviour in software first. The team should check that its stack avoids ROS 2, which the home page marks as unsupported.
Getting started with Drake
The quickest route listed is pip. Create a virtual environment and run pip install drake[mosek], leaving out the mosek extra if you have no MOSEK licence. The pip page requires pip 20.3 or later and asks macOS users to use Homebrew Python rather than Apple's system Python.
APT, Binary Download and Build From Source are the other options, and the binary releases do not include the Gurobi solver, so Gurobi needs a source build. Tutorials are Jupyter notebooks that can be run locally once the pip package is installed. The installation page lists tested configurations and Python versions from 3.12 to 3.15 depending on the row, so check yours against it first.
The case for and against Drake
- Drake grew out of MIT's Robot Locomotion Group and the Toyota Research Institute Robotics Division, and the repository applies the BSD 3-Clause licence to all components.
- Marked in our records as control design.
- pip and apt installs is among the traits we single out for it.
- BSD 3-Clause, one of the three traits the catalogue lists for Drake.
- Best for: Controls engineers who want analytical gradients and planning tools in Python.
- Binary releases leave out the Gurobi solver, so using it requires building from source.
- Poor match for: Teams that need a ROS 2 bridge or a graphical scene editor.
- ROS 2 and Julia are both marked unsupported on the home page.
What we make of Drake
Drake grew out of MIT's Robot Locomotion Group and the Toyota Research Institute Robotics Division, and the repository applies the BSD 3-Clause licence to all components. It models robot dynamics including friction and contact, and exposes equation structure such as sparsity and analytical gradients. Its ROS 2 entry reads unsupported, which matters to teams whose stack already runs on it.
Drake is entry #7 of 13 in the robotics simulation software list: MuJoCo sits directly above it and PyBullet directly below. After the challenger places, Drake is one of the established names, ranked in order and without a score.
The vendor states no country of establishment, so Drake is left out of the regional count for this list.
Drake pros and cons in practice
- The BSD 3-Clause licence applies and the pages show no price.
- Equation structure such as gradients and sparsity is exposed for algorithm writers.
- Tutorials are Jupyter notebooks that run locally.
- Four install routes are documented with tested configurations.
- The home page marks ROS 2 and Julia as unsupported.
- The binary releases lack the Gurobi solver, which needs a source build.
- Wheels require pip 20.3 or later and macOS users must use Homebrew Python.
- The pages we read mention no paid support route.
Paying for Drake
Charging model: BSD 3-Clause licensed; plans and prices do not appear in the documentation read.
The pages we read do not show any pricing or plan details. The software is distributed under the BSD 3-Clause licence in the RobotLocomotion/drake repository, whose text states that all components of Drake are licensed under it, with some portions subject to other permissive licences noted in the source. No paid support, hosted edition or consulting is mentioned. Third-party solvers can add costs: the pip instructions refer to a MOSEK licence for the mosek extra, and the installation page says the Gurobi solver is not in the binary releases. Those solvers are sold by their own vendors, and their prices are outside the pages we read.
Basis: drake.mit.edu, read for the Best Robotics Simulation Software in 2026 guide (reviewed 15 November 2026).
Living with Drake: setup and exit
Drake is organised around Systems, Solvers and Multibody APIs, installed through pip, apt, a binary download or a source build.
Wheels need pip 20.3 or later and nightly builds exist for Linux and macOS.
Continuous integration covers Ubuntu 24.04 and 26.04 on x86-64 and arm64 and macOS on arm64.
Tutorials are Jupyter notebooks that run locally, while integrations named include Python and LCM.
Drake: data and security
Drake is installed on your own hardware through pip, APT, a binary download or a source build, so models and results stay with you. The source is on GitHub in the RobotLocomotion organisation under the BSD 3-Clause licence, which lets a security team review it.
Wheel packages for Linux and macOS are published nightly, which means users should choose between dated nightly builds and releases deliberately. The pages we read state no certifications, audit reports or data processing terms. The project was started at MIT and Toyota Research Institute, and no country of establishment is stated.
Fact sheet: Drake
| Product | Drake |
|---|---|
| Vendor website | drake.mit.edu |
| Catalogue tags | BSD 3-Clause, Control design, pip and apt installs |
| Head-to-head pages | Not paired so far |
| What we do not hold | headcount, funding and customer count |
| Price visibility | Model recorded, no figure |
| Rank there | #7 of 13 |
| Primary guide | Best Robotics Simulation Software in 2026 |
| Alternatives page | None yet |
| Headquarters | Not stated by the vendor |
Drake integrations and ecosystem
Integrations the vendor names (10):
- Python
- C++
- Jupyter
- LCM
- MOSEK
- Gurobi (source build)
- pip
- APT
- GitHub
- Amazon Linux 2023
Also in our records:
- Drake is organised around Systems, Solvers and Multibody APIs, installed through pip, apt, a binary download or a source build.
- Continuous integration covers Ubuntu 24.04 and 26.04 on x86-64 and arm64 and macOS on arm64.
- Tutorials are Jupyter notebooks that run locally, while integrations named include Python and LCM.
What to look at instead of Drake
Shared tags and a similar price model decide who is listed here; where none match, the neighbours in the ranking fill in.
- RoboDK (#3 of 13): Not for: mobile robot, legged robot or learning-based research teams. Offline programming and simulation for industrial robot arms, priced per licence.
- Choreonoid (#4 of 13): MIT-licensed robotics GUI framework with a simulator and motion editor. Suits: Teams building their own robot tools wanting a permissive licence and a motion editor.
- Gazebo (#5 of 13): Apache-licensed 3D simulator maintained by Open Robotics for ROS users. Best for: ROS 2 teams on Ubuntu wanting a permissively licensed simulator with ready sensor models.
- ManiSkill (#9 of 13): GPU-parallel simulation framework for robot manipulation and learning.
- NVIDIA Isaac Sim (#11 of 13): Open source Omniverse reference framework for robot simulation and synthetic data. Best for: Teams generating synthetic training data on NVIDIA hardware with USD-based scenes.
Also ranked in
| Guide | Position |
|---|---|
| Best Robotics Simulation Software in 2026 | #7 of 13 |
Drake FAQ
9 answersIs Drake one of the top robotics simulation software?
In our robotics simulation software list Drake stands at place 7 of 13.
What else should I look at besides Drake?
Start with RoboDK, Choreonoid, Gazebo, ManiSkill and NVIDIA Isaac Sim. Each of them has its own entry in the same guide.
Who should choose Drake?
Controls engineers who want analytical gradients and planning tools in Python. It does not fit as well for teams that need a ROS 2 bridge or a graphical scene editor.
How much does Drake cost?
The pricing entry for Drake reads: BSD 3-Clause licensed; plans and prices do not appear in the documentation read. On free access, we recorded no free plan.
Which country is Drake from?
The vendor pages we reviewed state no country of establishment for Drake, so we do not give one.
Who started Drake?
The credits page says the project was started by members of the Robot Locomotion Group at MIT and the Robotics Division at Toyota Research Institute.
Which licence does it use?
BSD 3-Clause, stated in the RobotLocomotion/drake repository, with some portions under other permissive licences.
Does it work with ROS 2?
The home page marks ROS 2 and Julia as unsupported, and lists Python and LCM as integrations.
How do I install it?
Through pip, APT, a binary download or a source build. The pip route needs pip 20.3 or later.
Drake: sources
Read from the vendor’s own pages on 14 November 2026.
- drake.mit.edu: Toolbox description, APIs, equation structure, integrations, tutorials, friction and contact, no pricing.
- drake.mit.edu/pip.html: Pip install, mosek extra, pip 20.3, Homebrew Python, nightly wheels.
- drake.mit.edu/installation.html: Tested configurations, Python versions, Gurobi source build.
- drake.mit.edu/credits.html: Started by MIT Robot Locomotion Group and Toyota Research Institute.
- raw.githubusercontent.com/RobotLocomotion/drake/master/LICENSE.TXT: BSD 3-Clause licence for all components.
How we rank
Four tests decide a position: how soon the product is useful, what the bill really becomes, how data leaves and who it suits. The result for Drake: #7, out of 13 ranked. Details: how we rank software.