# New Release: Ultralytics v8.4.70 **URL:** \ **Category:** Discussion\ **Tags:** ultralytics-official, releases, announcements\ **Created:** [June 17, 2026, 2:12pm UTC](https://community.ultralytics.com/t/new-release-ultralytics-v8-4-70/2071 "2026-06-17T14:12:13Z")\ **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:** [June 17, 2026, 2:12pm UTC](https://community.ultralytics.com/t/new-release-ultralytics-v8-4-70/2071/1 "2026-06-17T14:12:13Z")
# Ultralytics v8.4.70 is out πŸŽ‰ ## Summary Ultralytics v8.4.70 is a stability and usability release focused on reliability πŸ›  The headline fix improves **multi-GPU training consistency** by correctly synchronizing AMP state across DDP ranks, alongside **safer checkpoint loading** , a **segmentation plotting fix** , and several **helpful documentation improvements** for Ultralytics YOLO and [Ultralytics Platform](https://platform.ultralytics.com). If you’re training on multiple GPUs, working with segmentation outputs, or deploying newer YOLO26 workflows, this release is especially worth updating to βœ… ## Highlights - πŸš€ **Critical fix for DDP AMP synchronization in multi-GPU training** - πŸ›‘ **Clearer, safer checkpoint loading errors** - 🎨 **Fixed `Results.plot()` for masks-only segmentation outputs** - πŸ”₯ **Improved YOLO26 heatmaps docs with real tracking use cases** - πŸ“ **Better training docs for `save_dir` output control** ## New Features and Improvements ### πŸš€ Multi-GPU training fix The most important change in v8.4.70 is [PR #24810](https://github.com/ultralytics/ultralytics/pull/24810) by [@LZ-QWQ](https://github.com/LZ-QWQ), which fixes **DDP AMP broadcast state synchronization**. Previously, during distributed multi-GPU training, **AMP mixed precision could be disabled on one GPU rank while remaining enabled on others**. That mismatch could lead to unstable or confusing behavior. With this fix, all ranks now correctly share the same AMP setting. This is a big reliability improvement for users training Ultralytics YOLO models across multiple GPUs. ### πŸ›‘ Safer checkpoint loading [PR #24857](https://github.com/ultralytics/ultralytics/pull/24857) by [@glenn-jocher](https://github.com/glenn-jocher) improves checkpoint loading by raising **clear, user-friendly errors** for unsupported formats, including plain `state_dict` files and some older YOLOv5-style module references. Instead of cryptic failures like `AttributeError` or `KeyError`, users now get much clearer feedback about what went wrong. ### 🎨 Segmentation plotting fix [PR #24635](https://github.com/ultralytics/ultralytics/pull/24635) by [@ahmet-f-gumustas](https://github.com/ahmet-f-gumustas) fixes an `AttributeError` in `Results.plot()` when handling **masks-only results**. This makes segmentation workflows more robust, especially for custom result objects or pipelines that don’t include bounding boxes. ## Documentation Updates ### πŸ”₯ YOLO26 heatmaps guide refresh [PR #24815](https://github.com/ultralytics/ultralytics/pull/24815) by [@raimbekovm](https://github.com/raimbekovm) rewrites the heatmaps guide with a stronger focus on **tracking and video analytics**. The updated docs now better explain practical YOLO26 heatmap use cases like: - traffic flow analysis - crowd movement monitoring - region-based counting - retail and transportation analytics ### πŸ“ `save_dir` training argument documented [PR #24831](https://github.com/ultralytics/ultralytics/pull/24831) by [@raimbekovm](https://github.com/raimbekovm) adds documentation for the `save_dir` training argument, making it easier to send training outputs to an exact folder without auto-increment behavior. ### 🧩 Quality-of-life doc and config fixes A set of smaller improvements also landed in this release: - [PR #24833](https://github.com/ultralytics/ultralytics/pull/24833) by [@raimbekovm](https://github.com/raimbekovm) removes the unused `region` argument from the instance segmentation guide - [PR #24832](https://github.com/ultralytics/ultralytics/pull/24832) by [@raimbekovm](https://github.com/raimbekovm) fixes `figsize` typing in `SolutionConfig` - [PR #24822](https://github.com/ultralytics/ultralytics/pull/24822) by [@raimbekovm](https://github.com/raimbekovm) fixes docs rendering issues - [PR #24826](https://github.com/ultralytics/ultralytics/pull/24826) by [@pderrenger](https://github.com/pderrenger) updates formatting, headers, and redirects - [PR #24856](https://github.com/ultralytics/ultralytics/pull/24856) by [@glenn-jocher](https://github.com/glenn-jocher) fixes the TensorRT quantization docs link - [PR #24858](https://github.com/ultralytics/ultralytics/pull/24858) by [@glenn-jocher](https://github.com/glenn-jocher) fixes docs workflow PR branch sync ## Why this release matters v8.4.70 is not a flashy feature release, but it delivers meaningful day-to-day improvements: - βœ… more reliable distributed training - βœ… fewer confusing checkpoint-loading failures - βœ… better segmentation result handling - βœ… clearer guidance for real-world YOLO26 workflows - βœ… smoother docs and configuration experience In short, this release makes Ultralytics YOLO more dependable for training, debugging, and deployment. ## New Contributor πŸ™Œ A big welcome to [@LZ-QWQ](https://github.com/LZ-QWQ), who made their first contribution with [PR #24810](https://github.com/ultralytics/ultralytics/pull/24810) β€” and it was an important one! ## Try it out You can upgrade with: ```bash pip install -U ultralytics ``` Then explore the release notes in the [v8.4.70 release page](https://github.com/ultralytics/ultralytics/releases/tag/v8.4.70) or review everything included in the [full changelog](https://github.com/ultralytics/ultralytics/compare/v8.4.69...v8.4.70). If you test the release, we’d love to hear how it works for your workflows, especially for **multi-GPU training** , **segmentation** , and **YOLO26 tracking or heatmap use cases** πŸ’¬