# New Release: Ultralytics v8.4.127 **URL:** \ **Category:** Discussion\ **Tags:** ultralytics-official, releases, announcements\ **Created:** [August 24, 2026, 1:21pm UTC](https://community.ultralytics.com/t/new-release-ultralytics-v8-4-127/2184 "2026-08-24T13:21:02Z")\ **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:** [August 24, 2026, 1:21pm UTC](https://community.ultralytics.com/t/new-release-ultralytics-v8-4-127/2184/1 "2026-08-24T13:21:02Z")
# 🚀 Ultralytics v8.4.127: More Reliable Exports, Training, and Deployment Ultralytics **v8.4.127** makes exported YOLO models reliably load with the correct task and model family across all 20 supported formats. This release also improves OpenVINO and CoreML deployment, training recovery, result handling, tracking, datasets, and documentation. Review the complete details on the [Ultralytics v8.4.127 release page](https://github.com/ultralytics/ultralytics/releases/tag/v8.4.127) before exploring the highlights below. ## 🌟 New Features - **Correct task detection across all 20 export formats:** [PR #25886](https://github.com/ultralytics/ultralytics/pull/25886) by [@artest08](https://github.com/artest08) now reads embedded export metadata instead of guessing from filenames or paths. This preserves segmentation masks and pose keypoints while correctly routing RT-DETR outputs through both the Python API and CLI. - **CoreML NMS for segmentation and pose:** [PR #25873](https://github.com/ultralytics/ultralytics/pull/25873) by [@SergioAlmeida29](https://github.com/SergioAlmeida29) adds `nms=True` support while keeping masks and keypoints aligned with retained boxes. 🍎 - **Direct access to official Platform datasets:** [PR #25892](https://github.com/ultralytics/ultralytics/pull/25892) by [@glenn-jocher](https://github.com/glenn-jocher) links all 52 publicly hosted datasets from the documentation, making it easier to preview, clone, annotate, train, and deploy through the [Ultralytics Platform](https://platform.ultralytics.com). ## ⚡ Improvements - [PR #25893](https://github.com/ultralytics/ultralytics/pull/25893) by [@glenn-jocher](https://github.com/glenn-jocher) applies the correct OpenVINO throughput hint for single-device dynamic batching. - [PR #25881](https://github.com/ultralytics/ultralytics/pull/25881) by [@Y-T-G](https://github.com/Y-T-G) documents YOLO26 Objects365 pretraining, COCO fine-tuning, embedded training logs, training curves, and checkpoint code revisions. 📚 - [PR #25856](https://github.com/ultralytics/ultralytics/pull/25856) by [@raimbekovm](https://github.com/raimbekovm) rewrites the YOLOE documentation and clarifies the evaluation protocol on both LVIS results pages. - [PR #25871](https://github.com/ultralytics/ultralytics/pull/25871) by [@raimbekovm](https://github.com/raimbekovm) selects the correct YOLOE model and validator by task in the shared trainer. - [PR #25891](https://github.com/ultralytics/ultralytics/pull/25891) by [@Rahulbiradar9](https://github.com/Rahulbiradar9) serializes custom augmentations as portable text representations in checkpoints, improving portability and reproducibility. - [PR #25888](https://github.com/ultralytics/ultralytics/pull/25888) by [@Kagura-Ahad](https://github.com/Kagura-Ahad) adds support for `Path` filenames in `imread`. - [PR #25166](https://github.com/ultralytics/ultralytics/pull/25166) by [@Y-T-G](https://github.com/Y-T-G) fits TQDM output to the available terminal width. - [PR #25900](https://github.com/ultralytics/ultralytics/pull/25900) by [@Y-T-G](https://github.com/Y-T-G) streams live progress-bar updates to console-log consumers. - [PR #25884](https://github.com/ultralytics/ultralytics/pull/25884) by [@raimbekovm](https://github.com/raimbekovm) moves documentation media to the CMS assets CDN for more reliable previews in tools such as Slack and Telegram. ## 🛠 Bug Fixes - **Safer training resume:** [PR #25816](https://github.com/ultralytics/ultralytics/pull/25816) by [@Ajaysingh-2003](https://github.com/Ajaysingh-2003) prevents custom `pretrained` weights from replacing checkpoint weights during resume, preserving the model, optimizer, scaler, EMA, epoch, and distillation state from `last.pt`. 🔄 - **Dense result preservation:** [PR #25880](https://github.com/ultralytics/ultralytics/pull/25880) by [@JESUSROYETH](https://github.com/JESUSROYETH) ensures semantic segmentation and depth indexing retains complete dense maps rather than returning only one image row. - **DeepOCSORT matching:** [PR #25875](https://github.com/ultralytics/ultralytics/pull/25875) by [@JESUSROYETH](https://github.com/JESUSROYETH) passes last-observation IoU into OCR appearance fusion, preventing incorrect matches between zero-overlap objects. - **TrackTrack behavior:** [PR #25877](https://github.com/ultralytics/ultralytics/pull/25877) by [@JESUSROYETH](https://github.com/JESUSROYETH) aligns new-track prediction and confirmation with the reference implementation. - **YOLOE training stability:** [PR #25867](https://github.com/ultralytics/ultralytics/pull/25867) by [@raimbekovm](https://github.com/raimbekovm) fixes prompt-free segmentation crashes during final validation and re-enables CLIP tests on Python 3.12. - **OpenVINO INT8 stability:** [PR #25898](https://github.com/ultralytics/ultralytics/pull/25898) by [@glenn-jocher](https://github.com/glenn-jocher) addresses dynamic-shape segmentation faults on Intel AMX CPUs and replaces a dead documentation link. ## 👋 Welcome, New Contributors A warm welcome to [@Ajaysingh-2003](https://github.com/Ajaysingh-2003), who made a first contribution through [PR #25816](https://github.com/ultralytics/ultralytics/pull/25816), and [@Kagura-Ahad](https://github.com/Kagura-Ahad), who joined through [PR #25888](https://github.com/ultralytics/ultralytics/pull/25888). Thank you for helping improve Ultralytics YOLO! 🎉 ## 📦 Try v8.4.127 Upgrade to the latest release with: ```bash pip install -U ultralytics ``` Whether you export specialized models, resume long training runs, deploy with OpenVINO or CoreML, or work with dense prediction tasks, v8.4.127 should provide a more dependable workflow. Explore every change in the [full v8.4.126 to v8.4.127 changelog](https://github.com/ultralytics/ultralytics/compare/v8.4.126...v8.4.127), try the release, and let the Ultralytics team and YOLO community know how it performs in your projects. 🚀