# New Release: Ultralytics v8.4.66

**URL:** <https://community.ultralytics.com/t/new-release-ultralytics-v8-4-66/2058>\
**Category:** Discussion\
**Tags:** ultralytics-official, releases, announcements\
**Created:** [June 12, 2026, 2:02pm UTC](https://community.ultralytics.com/t/new-release-ultralytics-v8-4-66/2058 "2026-06-12T14:02:42Z")\
**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:** [June 12, 2026, 2:02pm UTC](https://community.ultralytics.com/t/new-release-ultralytics-v8-4-66/2058/1 "2026-06-12T14:02:42Z")

</div>

## Ultralytics v8.4.66 is out 🎉

A new **Ultralytics YOLO** release has landed: **v8.4.66**. This update focuses on practical quality improvements across three important areas: **more reliable NVIDIA GPU monitoring** 🔋, **better semantic segmentation filtering and metrics** 🧠, and **sharper CoreML semantic exports** on iPhone and Apple devices 📱.

If you’re building with **[Ultralytics Platform](https://platform.ultralytics.com)** or working locally with **[YOLO26](https://platform.ultralytics.com/ultralytics/yolo26)**, this release should make monitoring, evaluation, and mobile deployment smoother without changing your core workflow.

* * *

## Quick summary

### Highlights in v8.4.66

- 🚀 More reliable built-in **NVIDIA GPU monitoring**
- 🎯 Proper `classes` filtering for **semantic segmentation**
- 📈 More accurate **semantic segmentation mIoU reporting**
- 📱 Full-resolution **CoreML semantic maps** for better iPhone results
- ⚙ Cleaner **TensorRT 11** export flow
- 📝 Clearer docs across **YOLO26** testing and training guides
- 🛠 More robust CI and packaging setup

* * *

## New features

### 🔋 More reliable NVIDIA GPU monitoring

GPU stats collection is now more dependable thanks to direct use of NVIDIA’s NVML library.

- `nvidia-ml-py` is now included as a standard dependency
- Built-in GPU reporting no longer relies on runtime install attempts
- Utilities like auto device selection and system logging should behave more consistently on NVIDIA systems

This improvement comes from [PR #23922 by @glenn-jocher](https://github.com/ultralytics/ultralytics/pull/23922) from [@glenn-jocher](https://github.com/glenn-jocher).

### 🎯 Better semantic segmentation class filtering

Semantic segmentation now handles the `classes` filter much more intuitively across the workflow.

This includes:

- dataset loading
- prediction
- validation
- metric reporting

Non-selected classes are now properly ignored instead of leaking into results, and unsupported configurations like `single_cls=True` for semantic segmentation are handled more safely.

This feature was added in [PR #24806 by @lmycross](https://github.com/ultralytics/ultralytics/pull/24806) from [@lmycross](https://github.com/lmycross).

### 📱 Full-resolution CoreML semantic exports

CoreML exports for semantic segmentation now preserve **full-resolution semantic maps** rather than lower-resolution stride-8 maps.

That means:

- sharper overlays
- less blocky masks
- better visual quality on iPhone and other Apple devices

This landed in [PR #24799 by @glenn-jocher](https://github.com/ultralytics/ultralytics/pull/24799) from [@glenn-jocher](https://github.com/glenn-jocher).

* * *

## Improvements

### 📈 More accurate semantic segmentation metrics

Semantic segmentation evaluation is now more faithful to the actual data being evaluated.

- mIoU now ignores classes not present in the ground truth
- per-class summaries better reflect what was truly measured

This makes reported results more trustworthy, especially when working with filtered or sparse class sets.

### ⚙ Cleaner TensorRT 11 export path

TensorRT 11 export was simplified for cleaner precision handling.

- removed the separate `half` argument from ONNX precision conversion
- simplified FP16 and INT8 export flow
- streamlined INT8 calibration preprocessing

This improvement comes from [PR #24795 by @Laughing-q](https://github.com/ultralytics/ultralytics/pull/24795) from [@Laughing-q](https://github.com/Laughing-q).

### 📊 Refreshed iPhone CoreML benchmarks

The iPhone CoreML benchmark docs were updated using optimized profile-mode measurements, giving a better picture of real **YOLO26** CoreML behavior, especially for semantic models.

See [PR #24804 by @glenn-jocher](https://github.com/ultralytics/ultralytics/pull/24804) from [@glenn-jocher](https://github.com/glenn-jocher).

* * *

## Documentation updates

This release also includes several documentation improvements to make testing, training, and integrations easier to follow.

- Improved model testing guide structure and corrected validation metric claims in [PR #24800 by @raimbekovm](https://github.com/ultralytics/ultralytics/pull/24800) from [@raimbekovm](https://github.com/raimbekovm)
- Improved **YOLO26** training recipe guide structure and accuracy in [PR #24805 by @raimbekovm](https://github.com/ultralytics/ultralytics/pull/24805) from [@raimbekovm](https://github.com/raimbekovm)
- Fixed the JupyterLab evaluate link target in [PR #24808 by @raimbekovm](https://github.com/ultralytics/ultralytics/pull/24808) from [@raimbekovm](https://github.com/raimbekovm)
- Clarified multiple bounding box behavior in SAM 3 semantic predictor docs in [PR #24807 by @raimbekovm](https://github.com/ultralytics/ultralytics/pull/24807) from [@raimbekovm](https://github.com/raimbekovm)

* * *

## Bug fixes and maintenance

### 🛠 CI reliability improvements

To reduce flaky CI failures, dependency installs now automatically retry.

That change is included in [PR #24809 by @glenn-jocher](https://github.com/ultralytics/ultralytics/pull/24809) from [@glenn-jocher](https://github.com/glenn-jocher).

### 📦 Packaging cleanup

Because `nvidia-ml-py` is now a standard dependency, extra manual installs were removed from CI and Docker-related setup, helping keep environments cleaner and more predictable.

* * *

## Why this release matters

**v8.4.66** is a quality-focused release with real day-to-day impact:

- ✅ fewer GPU monitoring issues on NVIDIA systems
- 🎨 better usability for semantic segmentation workflows
- 📏 more trustworthy segmentation metrics
- 🍎 sharper mobile semantic outputs with CoreML
- ⚡ simpler TensorRT 11 export behavior
- 📚 clearer guidance for testing and training with YOLO26

If you’re starting a new project, **[YOLO26](https://platform.ultralytics.com/ultralytics/yolo26)** remains the recommended choice. **[YOLO11](https://platform.ultralytics.com/ultralytics/yolo11)** is still fully supported, but YOLO26 is smaller, faster, more accurate, and natively end-to-end.

* * *

## Full changelog

You can review the complete diff in the [full changelog for v8.4.66](https://github.com/ultralytics/ultralytics/compare/v8.4.65...v8.4.66), and the packaged release is available on the [v8.4.66 release page](https://github.com/ultralytics/ultralytics/releases/tag/v8.4.66).

* * *

## Try it out and let us know what you think 💬

We’d love your feedback on this release, especially if you’re using:

- semantic segmentation
- CoreML exports
- TensorRT 11 deployment
- NVIDIA-based training or benchmarking

Thanks to all contributors who helped ship **Ultralytics v8.4.66** 🙌