Open Source RoboticsRobots, software & learning

Taichi

Software

Python-embedded parallel programming language and JIT compiler for CPU/GPU numerical kernels, including physics simulation examples. [1]

Installation & first result →Official websiteGetting started

For
Developers implementing custom simulation or numerical computation [1]
Source rights
Named licenses are documented by part. Review licenses
Unknowns
operating system, compute requirements not fully established. Review gaps

Source researched · Not tested by us
Sources checked · Research details

Licenses & related hardware

Requirements · Research details and dates

On this page

Overview

Taichi is a Python-embedded language and just-in-time compiler for numerical kernels on CPUs and GPUs. It is a building block for custom physics and robotics computation rather than a ready-made robot-control environment. Its examples expose the numerical method and data fields directly. [1]

Is this for you?

Known limitations
  • Package metadata and README disagree about supported Python versions.
  • GPU preference can fall back to CPU unless fallback is disabled.
  • A material demo does not establish calibrated robotics contact accuracy.

Editorial assessment from cited documentation; no runtime testing.

[1] [2] [3] [4] · Editorial assessment

What you need

Dependencies
  • Release setup.py requires Python >=3.9,<4.0
  • Compatible platform wheel or a separately configured source build
  • Graphical display for interactive examples

Editorial assessment from cited documentation; no runtime testing.

[2] [3] · Editorial assessment

Installation & first result

Open the upstream getting-started guide [3]

  • Install the reviewed 1.7.4 package in an isolated environment.
  • Check out v1.7.4 and run the mpm128.py example described in the overview.
  • Inspect which backend initialized before interpreting performance.

Editorial assessment from cited documentation; no runtime testing.

[1] [3] [4] · Editorial assessment

Installation

Install taichi==1.7.4 in an isolated supported Python environment; use the matching release example sources.

Editorial assessment from cited documentation; no runtime testing.

[1] [6] · Editorial assessment

Hardware, models & data

Inputs and outputs

User-written kernels update numeric fields; the MPM example displays evolving simulated material particles.

[1] [3]

Project notes

Inspect a small material simulation

The reviewed release is v1.7.4, commit b4b956fdaa67fd3e198527fd607053ecb9bf4aa2. Install taichi==1.7.4 in an isolated environment with a compatible wheel, and use the matching tagged checkout for examples. The README's Python 3.6–3.10 statement is stale relative to setup.py, which declares Python 3.9 or newer and below 4.0. That metadata range alone does not establish wheel availability on every Python/platform combination. [1] [2] [6]

From the checkout, run python python/taichi/examples/simulation/mpm128.py. The example initializes 9,000 two-dimensional particles on a 128 × 128 grid of nodes at its default quality setting. It demonstrates material-point simulation with liquid, jelly and snow behaviors. A graphical window accepts keyboard changes to gravity and mouse attraction/repulsion; it is not a headless benchmark. [3]

The example requests ti.gpu, while initialization enables backend fallback by default. If none of the requested backends is supported, selection can fall back to CPU. Inspect the selected backend when comparing runs; a visible animation does not by itself prove GPU execution or a particular throughput. [3] [4]

Use the example to understand kernels, fields and material updates before adapting it to a robotics problem. Its default material parameters and interactive behavior are not calibrated evidence for real robot contact or deformation. Source is Apache-2.0, with dependencies retaining their own terms. No installation, numerical validation or performance benchmark was run in this review. [3] [7]

What we checked

Source researched only; no installation, training, benchmark or physical hardware test performed. No affiliate links or reused media.

Not yet established

  • Requirements: operating system, host computer
  • Version-specific hardware compatibility (ecosystem tags alone do not establish it)

Updates

Upstream last updated
Jul 30, 2025 [6]
Latest release
Jul 31, 2025 [6]
Profile updated
Oct 17, 2026
Sources last checked
Oct 17, 2026
Next review
Nov 17, 2026
Reviewed revision
v1.7.4 / b4b956fdaa67fd3e198527fd607053ecb9bf4aa2 · software repository

Source files and licenses

Numerical programming framework relevant to custom robotics simulation, with backend/version limits disclosed.

LayerAvailability / license
softwareavailable · Apache-2.0
Repository materials only; external components retain their own terms. [7]
documentationavailable · Apache-2.0
Repository materials only; external components retain their own terms. [7]

The material-point example uses no trained weights.

The example initializes particles procedurally.

References

Each source links to the exact revision we read. “Used for” shows where this page relies on it.

Correction guidance · Discussion is not enabled · Profile format 0.7.0-draft.5

Change history