New Release: Ultralytics v8.3.95

:rocket: Announcing Ultralytics v8.3.95 Release!

We are thrilled to announce the release of Ultralytics v8.3.95, packed with powerful new features, enhancements, and improvements designed to elevate the YOLO11 experience. Whether you’re training advanced models, deploying with GPUs, or leveraging integrations, this release brings something for everyone!


:glowing_star: Summary

The v8.3.95 release introduces a CUDA-optimized Dockerfile for seamless multi-GPU training and inference, updates key dependencies for improved compatibility, and polishes the user experience with significant documentation and CLI improvements. Dive into the details below!


:bar_chart: New Features

NVIDIA CUDA Dockerfile for YOLO11

We’ve added a brand-new Dockerfile-nvidia-cuda to bring you optimized GPU acceleration for training and inference workloads. Built on NVIDIA’s CUDA runtime (version: 12.8.1), this addition ensures unparalleled performance on high-end hardware.

Learn more about this enhancement and contribute through this PR by @glenn-jocher.


:counterclockwise_arrows_button: Improvements

Dependency Updates

Keeping Ultralytics cutting-edge, we’ve:

  • Upgraded PyTorch base image to 2.6.0-cuda12.6-cudnn9-runtime for enhanced compatibility and stability (details here).
  • Updated CoreML compatibility to version 8.0, ensuring support for the latest advancements (view update).

Documentation Enhancements

We’ve enhanced guides across the board with:

CLI Standardization

To maintain consistency, array inputs such as classes and kpts now require quotes in the CLI. This change improves usability and prevents common errors (PR reference by @RizwanMunawar).

GitHub Banner & YOLOE Guide Updates

  • GitHub now redirects directly to the latest YOLO11 blog for a richer user experience.
  • Updated informative tips and sections for the YOLOE documentation (PR overview).

:hammer_and_wrench: Bug Fixes

  • Improved bash codeblock formatting (PR by @glenn-jocher).
  • Addressed minor display issues in CLI and documentation tips.

:bullseye: Impact

Why You’ll Love This Update:

  • Supercharged Performance: GPU-enhanced Dockerfile streamlines demanding tasks effortlessly.
  • Improved Compatibility: Software updates future-proof Ultralytics, keeping you aligned with the latest tools.
  • Simplified Workflows: Standardized CLI syntax and enhanced documentation refine your experience.
  • Community-Driven: Each improvement reflects the incredible collaboration of contributors worldwide.

:light_bulb: How to Get Started

Ready to explore these updates? Simply update to the latest version with:

pip install ultralytics --upgrade

Or pull the new CUDA-enabled Dockerfile for GPU training:

docker pull ultralytics/nvidia-cuda:8.3.95

Find the full details of the release in our v8.3.95 Release Notes. For a complete change log, visit the compare page.


:raising_hands: Feedback & Discussion

Love the update? Have ideas or suggestions? Join our Discussions and share your thoughts. As always, your feedback drives innovation and helps make YOLO11 even better!


We extend our gratitude to all contributors and the amazing YOLO community for your dedication and feedback. Let’s continue building and innovating together!

Happy coding,
The Ultralytics Team