# New Release: Ultralytics v8.4.212

**URL:** <https://community.ultralytics.com/t/new-release-ultralytics-v8-4-212/2309>\
**Category:** Discussion\
**Tags:** releases, announcements, ultralytics-official\
**Created:** [November 14, 2026, 8:09am UTC](https://community.ultralytics.com/t/new-release-ultralytics-v8-4-212/2309 "2026-11-14T08:09:57Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**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:** [November 14, 2026, 8:09am UTC](https://community.ultralytics.com/t/new-release-ultralytics-v8-4-212/2309/1 "2026-11-14T08:09:57Z")

</div>

# 🚀 Ultralytics v8.4.212: Expanded OCR and More Reliable Workflows

**Ultralytics v8.4.212 expands OCR to more writing systems and improves model discovery, data imports, training reliability, and inference compatibility.** This release also brings safer label handling, richer validation exports, and deployment fixes across multiple runtimes.

## 🌍 New Features and Platform Updates

### OCR for more languages and scripts

Choose a **PP-OCRv5 recognizer** for Korean, Cyrillic, Arabic, Devanagari, Thai, Greek, Tamil, or Telugu text. These models run through **ONNX Runtime** , without requiring PaddlePaddle or PaddleOCR.

Select the recognizer that matches your expected text: OCR **does not automatically detect or switch languages** , and Hebrew is not included in these additions.

- [PR #27167](https://github.com/ultralytics/ultralytics/pull/27167) by [@Laughing-q](https://github.com/Laughing-q) adds the expanded OCR recognizer support.

### Easier model discovery and dataset imports

Ultralytics Platform documentation now covers the **Explore Models tab** , where you can search public models by what they detect, browse results, and try or clone models. Updated guides also explain direct imports from hosted Label Studio projects for detection, oriented-box detection, segmentation, and classification.

- [PR #27160](https://github.com/ultralytics/ultralytics/pull/27160) by [@pderrenger](https://github.com/pderrenger) documents Explore Models, HEIC previews in Predict, and dataset status in Recent Activity.
- [PR #27163](https://github.com/ultralytics/ultralytics/pull/27163) by [@glenn-jocher](https://github.com/glenn-jocher) refreshes VLM leaderboard picks, model deployment and ranking wording, and Predict screenshots.
- [PR #27176](https://github.com/ultralytics/ultralytics/pull/27176) by [@glenn-jocher](https://github.com/glenn-jocher) clarifies cloning to another workspace versus duplicating in place with Copy and Paste.
- [PR #27178](https://github.com/ultralytics/ultralytics/pull/27178) by [@pderrenger](https://github.com/pderrenger) documents direct Label Studio imports and image deletion confirmation.
- [PR #27180](https://github.com/ultralytics/ultralytics/pull/27180) by [@glenn-jocher](https://github.com/glenn-jocher) updates Label Studio and Roboflow import screenshots to show the actual storage limit.

## 🧰 Training and Validation Improvements

Training cancellation is more dependable, older macOS MPS configurations receive determinism fixes, and exported validation summaries include more task-specific metrics.

- [PR #27161](https://github.com/ultralytics/ultralytics/pull/27161) by [@artest08](https://github.com/artest08) fixes distributed training hangs when Platform cancels a run mid-epoch.
- [PR #27159](https://github.com/ultralytics/ultralytics/pull/27159) by [@raimbekovm](https://github.com/raimbekovm) restores deterministic semantic, depth, and segmentation training on MPS with PyTorch versions below 2.3.
- [PR #27162](https://github.com/ultralytics/ultralytics/pull/27162) by [@raimbekovm](https://github.com/raimbekovm) divides the dataloader worker cap by the number of GPUs in the run.
- [PR #27184](https://github.com/ultralytics/ultralytics/pull/27184) by [@raimbekovm](https://github.com/raimbekovm) adds per-class mask mAP, pose mAP, and semantic pixel accuracy to validation CSV and JSON exports.

## ⚡ Inference and Deployment Fixes

This release improves runtime compatibility and CPU efficiency while addressing several task-specific inference issues.

- [PR #27175](https://github.com/ultralytics/ultralytics/pull/27175) by [@Zephyr4772](https://github.com/Zephyr4772) keeps ONNX inference running on the GPU with CUDA 12 PyTorch installations.
- [PR #27182](https://github.com/ultralytics/ultralytics/pull/27182) by [@var1914](https://github.com/var1914) improves CPU ONNX prediction speed on hyperthreaded machines and in CPU-limited containers.
- [PR #27185](https://github.com/ultralytics/ultralytics/pull/27185) by [@glenn-jocher](https://github.com/glenn-jocher) avoids a redundant 3 GB TensorRT download on JetPack 7, DGX, and NVIDIA containers.
- [PR #27177](https://github.com/ultralytics/ultralytics/pull/27177) by [@raimbekovm](https://github.com/raimbekovm) fixes incorrect MNN detections with `device=mps` or permuted tensor inputs.
- [PR #27171](https://github.com/ultralytics/ultralytics/pull/27171) by [@raimbekovm](https://github.com/raimbekovm) fixes YOLOE tracking with native ReID features.
- [PR #27181](https://github.com/ultralytics/ultralytics/pull/27181) by [@raimbekovm](https://github.com/raimbekovm) corrects misplaced Edge TPU pose keypoints for models with x, y-only keypoints, such as tiger-pose.

## 🛡 Safer, More Efficient Data Handling

- [PR #27170](https://github.com/ultralytics/ultralytics/pull/27170) by [@aswanth-07](https://github.com/aswanth-07) avoids unnecessary full-image copies when saving detection crops.
- [PR #27172](https://github.com/ultralytics/ultralytics/pull/27172) by [@mh-masum6875](https://github.com/mh-masum6875) ensures source manifests are read as UTF-8.
- [PR #27174](https://github.com/ultralytics/ultralytics/pull/27174) by [@AhmetMert01](https://github.com/AhmetMert01) prevents segmentation-mask conversion from overwriting label files with colliding filenames.

## 📚 Documentation and Community

- [PR #27165](https://github.com/ultralytics/ultralytics/pull/27165) by [@RizwanMunawar](https://github.com/RizwanMunawar) updates the video link in the segmentation object-isolation guide.
- [PR #27169](https://github.com/ultralytics/ultralytics/pull/27169) by [@raimbekovm](https://github.com/raimbekovm) removes outdated DeepSparse documentation following the project’s end of life.

🎉 A special welcome to **@AhmetMert01** for their first contribution, and thank you to everyone who helped improve this release!

## 🚀 Upgrade and Share Your Feedback

Install the release with:

```bash
pip install --upgrade ultralytics==8.4.212

```

Explore the [v8.4.212 release notes](https://github.com/ultralytics/ultralytics/releases/tag/v8.4.212) for the release overview, or review the [full changelog](https://github.com/ultralytics/ultralytics/compare/v8.4.211...v8.4.212) for all changes.

**Give v8.4.212 a try and let us know how it works for you!** Share your OCR results, deployment experiences, or any issues in the replies below. 💬
