# New Release: Ultralytics v8.4.95 **URL:** \ **Category:** Discussion\ **Tags:** releases, announcements, ultralytics-official\ **Created:** [July 14, 2026, 1:33pm UTC](https://community.ultralytics.com/t/new-release-ultralytics-v8-4-95/2111 "2026-07-14T13:33:42Z")\ **Posts on this page:** 1\ **Page:** 1
**Author:** ![glenn-jocher](https://sea1.discourse-cdn.com/flex001/user_avatar/community.ultralytics.com/glenn-jocher/32/83_2.png) [@glenn-jocher](https://community.ultralytics.com/u/glenn-jocher)\ **Post date:** [July 14, 2026, 1:33pm UTC](https://community.ultralytics.com/t/new-release-ultralytics-v8-4-95/2111/1 "2026-07-14T13:33:42Z")
# 🚀 Ultralytics v8.4.95 Released Ultralytics **v8.4.95** improves training reliability, RT-DETR deployment behavior, image input compatibility, dataset efficiency, and documentation. Highlights include safer checkpoint loading for Ultralytics Platform GPU jobs, consistent `max_det` handling, broader NumPy image support, and COCO downloads reduced by roughly **7 GB**. 🌟 ## 🔒 Training and Input Reliability - **Safer checkpoint loading:** Raw training checkpoints no longer include training-only loss objects or NumPy scalar metadata that may be rejected by secure `weights_only=True` loading. This was completed in [PR #25154](https://github.com/ultralytics/ultralytics/pull/25154) by [@glenn-jocher](https://github.com/glenn-jocher). - **Expanded NumPy compatibility:** Grayscale, gray-plus-alpha, BGR, and BGRA inputs are normalized to the model’s expected channel count in `LoadPilAndNumpy`, thanks to [PR #25151](https://github.com/ultralytics/ultralytics/pull/25151) from [@JESUSROYETH](https://github.com/JESUSROYETH). - **Correct `uint8` tensor preprocessing:** Integer image tensors are now normalized without invalid floating-point dtype operations, while existing floating-point behavior remains unchanged. See [PR #23743](https://github.com/ultralytics/ultralytics/pull/23743) by [@jahsef](https://github.com/jahsef). ## 🎯 RT-DETR and Metrics Fixes - **Consistent RT-DETR `max_det` support:** Native prediction, validation, and decoder exports—including CoreML and other deployment formats—now correctly respect the configured detection limit through [PR #25153](https://github.com/ultralytics/ultralytics/pull/25153) by [@Hasnaathussain](https://github.com/Hasnaathussain). - **Correct pose metric labels:** `PoseMetrics.curves` no longer returns duplicated box labels, improving compatibility with experiment dashboards such as Weights & Biases. This fix arrived in [PR #25148](https://github.com/ultralytics/ultralytics/pull/25148) from [@JESUSROYETH](https://github.com/JESUSROYETH). ## 📦 Dataset and Conversion Improvements - **Smaller default COCO downloads:** COCO detection, segmentation, and pose configurations no longer download the optional 7 GB `test2017.zip` archive by default. Standard downloads are now approximately **20.2–20.3 GB** , following [PR #25136](https://github.com/ultralytics/ultralytics/pull/25136) by [@raimbekovm](https://github.com/raimbekovm). - **Deterministic segmentation conversion:** Mask files are processed in sorted filename order for more reproducible conversion and debugging. See [PR #24390](https://github.com/ultralytics/ultralytics/pull/24390) by [@AlexRTer](https://github.com/AlexRTer). - **Simpler empty-segment handling:** `segment2box` now immediately returns a valid zero box for empty segmentation inputs through [PR #23770](https://github.com/ultralytics/ultralytics/pull/23770) from [@1548093028](https://github.com/1548093028). ## 📚 Documentation and Examples - The `ObjectCounter` example now uses its returned results object correctly, thanks to [PR #23417](https://github.com/ultralytics/ultralytics/pull/23417) by [@parthivdholaria](https://github.com/parthivdholaria). - Training documentation now explains automatic dataset download behavior more precisely through [PR #24155](https://github.com/ultralytics/ultralytics/pull/24155) by [@vedantparnaik](https://github.com/vedantparnaik). - Prediction documentation and type hints now clarify that streaming returns an iterator while non-streaming calls return lists, following [PR #24391](https://github.com/ultralytics/ultralytics/pull/24391) by [@thiliapr](https://github.com/thiliapr). - Embedding return types received the same streaming clarification in [PR #24134](https://github.com/ultralytics/ultralytics/pull/24134) from [@waketzheng](https://github.com/waketzheng). ## 👋 Welcome, New Contributors! A warm welcome to first-time contributors [@thiliapr](https://github.com/thiliapr), [@parthivdholaria](https://github.com/parthivdholaria), [@jahsef](https://github.com/jahsef), [@vedantparnaik](https://github.com/vedantparnaik), [@1548093028](https://github.com/1548093028), [@AlexRTer](https://github.com/AlexRTer), [@waketzheng](https://github.com/waketzheng), and [@Hasnaathussain](https://github.com/Hasnaathussain). Thank you for helping improve Ultralytics YOLO! 🙌 ## ⬆ Try v8.4.95 Upgrade to the latest release with: ```bash pip install -U ultralytics ``` This release focuses on reliability, compatibility, exports, datasets, and documentation rather than introducing a new model family. **YOLO26 remains the latest stable and recommended Ultralytics model for new projects.** Explore the [v8.4.95 GitHub release](https://github.com/ultralytics/ultralytics/releases/tag/v8.4.95) for release assets, or review the [complete v8.4.94-to-v8.4.95 changelog](https://github.com/ultralytics/ultralytics/compare/v8.4.94...v8.4.95) for every change. Please give v8.4.95 a try and share your feedback, questions, or reproducible issues with the community. 🚀