nvidia-cutlass-dsl-libs-cu12
NVIDIA CUTLASS Python DSL
Decision gist · record as of 2026-08-14
Yes, if you have a compatible NVIDIA GPU (Ampere/Hopper/Blackwell), CUDA 12 environment, and need to write optimized CUDA kernels in Python. The active maintenance, strong upstream signal, and zero known vulnerabilities support adoption. However, verify the unclear license terms for your use case, and note that the package is in beta—expect potential API changes before summer 2026.AI-flagged interpretation of the facts on this page — verify before relying
Before you install
- Requires CUDA 12, Linux (x86_64 or aarch64), Python ≥3.10, and NVIDIA GPU with Ampere, Hopper, or Blackwell architecture.
- Medium install friction due to platform-specific wheels (x86_64 and aarch64 Linux only, Python 3.10–3.14) and seven runtime dependencies including cuda-python and nvidia-cuda-nvdisasm.
- Active maintenance with recent release (9 days old) and strong upstream repository signal (10250 stars).
License · maintenance · safety
(unclear) — License treatment is unclear; no SPDX identifier or raw license text is available. Verify licensing terms before use in proprietary or commercial contexts.
last release 2026-08-05 (9 days) · last repo commit 2026-08-14 · 10,250 stars
0 known vulnerabilities (OSV.dev, 2026-08-14) · 1,523,250 downloads/mo, #3,819 on PyPI
Alternatives
Verify before relying
pip install nvidia-cutlass-dsl-libs-cu12
import numpy
# Write kernels targeting Tensor Cores using CuTe DSL concepts- Specific API surface and programming model details beyond CuTe DSL concepts (layouts, tensors, hardware atoms).
- Performance benchmarks or comparative compile-time improvements claimed in the description.
- Beta stability guarantees and breaking-change policy before summer 2026 graduation.
- Integration patterns with deep-learning frameworks mentioned in description.
What it is and what it does
nvidia-cutlass-dsl-libs-cu12 is NVIDIA's Python interface to CUTLASS 4.x, a library for writing optimized CUDA kernels without deep C++ expertise. It exposes CuTe DSL, a low-level programming model that maps directly to CuTe C++ abstractions—layouts, tensors, hardware atoms, and thread/data hierarchy—enabling developers to target NVIDIA's Tensor Cores on modern GPUs (Ampere, Hopper, Blackwell) with full control over kernel structure.
The package is currently in public beta (graduating by summer 2026) and is designed to flatten the GPU programming learning curve, accelerate kernel prototyping, and integrate natively with deep-learning frameworks. It requires CUDA 12, runs on Linux (x86_64 and aarch64), and depends on numpy, typing-extensions, cuda-python, protobuf, and other NVIDIA runtime libraries.
Use it for
- Rapidly prototype and optimize matrix-multiply kernels for Tensor Cores without writing C++ code.
- Integrate custom CUDA kernels directly into deep-learning framework workflows using Python.
- Learn GPU programming and CuTe concepts with a lower barrier to entry than C++ CUTLASS.
- Develop high-throughput linear algebra operations targeting Ampere, Hopper, or Blackwell GPUs.
- Experiment with kernel designs and performance tuning in a native Python environment.
Worth the install?
AI-flagged interpretation of the facts on this page. Verify before relying on it.
Yes, if you have a compatible NVIDIA GPU (Ampere/Hopper/Blackwell), CUDA 12 environment, and need to write optimized CUDA kernels in Python.
The active maintenance, strong upstream signal, and zero known vulnerabilities support adoption. However, verify the unclear license terms for your use case, and note that the package is in beta—expect potential API changes before summer 2026.
Install
nvidia-cutlass-dsl-libs-cu12 on PyPI
Before you install
Medium install friction due to platform-specific wheels (x86_64 and aarch64 Linux only, Python 3.10–3.14) and seven runtime dependencies including cuda-python and nvidia-cuda-nvdisasm. Active maintenance with recent release (9 days old) and strong upstream repository signal (10250 stars).
Requires CUDA 12, Linux (x86_64 or aarch64), Python ≥3.10, and NVIDIA GPU with Ampere, Hopper, or Blackwell architecture.
License in practice
License treatment is unclear; no SPDX identifier or raw license text is available. Verify licensing terms before use in proprietary or commercial contexts.
Quickstart
pip install nvidia-cutlass-dsl-libs-cu12
import numpy
# Write kernels targeting Tensor Cores using CuTe DSL conceptsVerify before relying
- Specific API surface and programming model details beyond CuTe DSL concepts (layouts, tensors, hardware atoms).
- Performance benchmarks or comparative compile-time improvements claimed in the description.
- Beta stability guarantees and breaking-change policy before summer 2026 graduation.
- Integration patterns with deep-learning frameworks mentioned in description.
Package facts
| License | Not declared unclear |
| Python support | Supports the current Python release >=3.10 |
| Install friction | Medium. Platform-specific wheel |
| Runtime dependencies | 7 packagesnumpytyping-extensionscuda-pythonbackports.strenumprotobufnvidia-cuda-nvdisasmnvidia-cutlass-dsl-libs-base |
| Maintenance | Actively maintained 9 days since the last release |
| Last repo commit | |
| First released | |
| Downloads | 1,523,250 / month, #3,819 on PyPI 30-day window, as of 2026-08-14 |
| Known vulnerabilities | None known OSV.dev, checked 2026-08-14 |
| Classifiers | Development Status :: 4 - BetaEnvironment :: GPU :: NVIDIA CUDA :: 13License :: Other/Proprietary LicenseOperating System :: POSIX :: LinuxProgramming Language :: Python :: 3 :: OnlyProgramming Language :: Python :: 3.10Programming Language :: Python :: 3.11Programming Language :: Python :: 3.12Programming Language :: Python :: 3.13Programming Language :: Python :: 3.14Programming Language :: Python :: Implementation :: CPython |
Evidence: nvidia_cutlass_dsl_libs_cu12-4.7.0-cp310-cp310-manylinux_2_28_aarch64.whl; nvidia_cutlass_dsl_libs_cu12-4.7.0-cp310-cp310-manylinux_2_28_x86_64.whl; nvidia_cutlass_dsl_libs_cu12-4.7.0-cp311-cp311-manylinux_2_28_aarch64.whl; nvidia_cutlass_dsl_libs_cu12-4.7.0-cp311-cp311-manylinux_2_28_x86_64.whl; nvidia_cutlass_dsl_libs_cu12-4.7.0-cp312-cp312-manylinux_2_28_aarch64.whl; nvidia_cutlass_dsl_libs_cu12-4.7.0-cp312-cp312-manylinux_2_28_x86_64.whl; nvidia_cutlass_dsl_libs_cu12-4.7.0-cp313-cp313-manylinux_2_28_aarch64.whl; nvidia_cutlass_dsl_libs_cu12-4.7.0-cp313-cp313-manylinux_2_28_x86_64.whl; nvidia_cutlass_dsl_libs_cu12-4.7.0-cp314-cp314-manylinux_2_28_aarch64.whl; nvidia_cutlass_dsl_libs_cu12-4.7.0-cp314-cp314-manylinux_2_28_x86_64.whl; nvidia_cutlass_dsl_libs_cu12-4.7.0-cp314-cp314t-manylinux_2_28_aarch64.whl; nvidia_cutlass_dsl_libs_cu12-4.7.0-cp314-cp314t-manylinux_2_28_x86_64.whl
Tags
Let your AI agent find packages like this
Example. Real query, live index.
You found this page by searching. An agent finds it by wishing: SkillFed indexes 14,416 PyPI packages by what they can do, searchable in plain language.
wish › “CuTe DSL tensor cores”
- nvidia-cutlass-dsl-libs-cu12Provides a Python DSL for writing high-performance CUDA kernels using…
- nvidia-cutlass-dsl-libs-coreProvides a Python interface for writing high-performance CUDA kernels…
- nvidia-cutlass-dsl-libs-baseProvides Python interfaces for writing high-performance CUDA kernels…
Give your agent the search over MCP, or paste the wish link into any chat.
More Scientific/Engineering packages
NumPy provides an N-dimensional array object and a comprehensive suite of mathematical, linear algebra, Fourier transform, and random number functions for scientific computing in Python.
pandas provides fast, flexible data structures (Series and DataFrame) for loading, cleaning, transforming, and analyzing labeled or relational data in Python.
scipy provides numerical algorithms for mathematics, science, and engineering—including optimization, integration, linear algebra, Fourier transforms, signal and image processing, and ODE solvers—built on numpy arrays.
scikit-learn provides a comprehensive Python library for supervised and unsupervised machine learning, including classification, regression, clustering, dimensionality reduction, and model evaluation tools built on NumPy and SciPy.
Install it if you need to train, evaluate, or deploy supervised or unsupervised learning models.
dill extends Python's pickle module to serialize and deserialize a much wider range of Python objects, including functions, lambdas, classes, and interpreter sessions, to byte streams for storage or network transmission.
Multiprocess is an enhanced fork of Python's standard multiprocessing library that uses dill for better serialization, allowing you to spawn processes with a threading-like API and share complex objects between them.
Install it if you use multiprocessing and encounter pickle serialization limits with lambdas or complex objects.
See also nvidia-cutlass-dsl · nvidia-cutlass-dsl-libs-base · nvidia-cutlass-dsl-libs-cu13 · nvidia-cutlass-dsl-libs-core · flydsl · cuda-tile · cutensor-cu13 · cutensor-cu12 · nvidia-cudnn-frontend · transformer-engine-cu12