# New Release: Ultralytics v8.4.83

**URL:** <https://community.ultralytics.com/t/new-release-ultralytics-v8-4-83/2090>\
**Category:** Discussion\
**Tags:** releases, announcements, ultralytics-official\
**Created:** [June 30, 2026, 1:37pm UTC](https://community.ultralytics.com/t/new-release-ultralytics-v8-4-83/2090 "2026-06-30T13:37:08Z")\
**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 30, 2026, 1:37pm UTC](https://community.ultralytics.com/t/new-release-ultralytics-v8-4-83/2090/1 "2026-06-30T13:37:08Z")

</div>

# 🚀 Ultralytics v8.4.83 is here

**Quick summary:** Ultralytics v8.4.83 introduces a major deployment upgrade with the new **LiteRT export format** , which now unifies the previous TFLite and TF.js workflows into a single path for mobile, edge, embedded, and browser deployments. This release also improves segmentation memory usage and speed, strengthens training checkpoint reliability, and adds several stability fixes across training and export.

If you’re building with **Ultralytics YOLO** , especially **[YOLO26](https://platform.ultralytics.com/ultralytics/yolo26)** for mobile, embedded, edge, or web deployment, this is a great release to try. You can explore the full release on [GitHub here](https://github.com/ultralytics/ultralytics/releases/tag/v8.4.83).

* * *

## 🌟 Highlights

### 🔄 New LiteRT export replaces legacy TFLite and TF.js workflows

The biggest change in v8.4.83 is the new `format="litert"` export path, added in [PR #22870 by @ambitious-octopus](https://github.com/ultralytics/ultralytics/pull/22870) and contributed by [@ambitious-octopus](https://github.com/ambitious-octopus).

With this update:

- `litert` becomes the new unified export format for `.tflite` deployment
- Legacy `tflite` and `tfjs` names still work, but now emit deprecation warnings and redirect to LiteRT
- Exported `.tflite` models can now be loaded directly inside Ultralytics for `predict` and `val`
- Browser deployment is simplified through LiteRT.js using the same `.tflite` model

This creates a much cleaner deployment story for teams shipping models across devices and platforms. 📱🌐

### 📦 Broader on-device deployment support

The new LiteRT export path supports multiple quantization modes, making deployment more flexible on constrained hardware:

- standard INT8
- mixed INT8 with 16-bit activations
- dynamic INT8 without calibration data

This is especially useful for production deployments of YOLO26 on mobile and edge devices.

### 🎭 Faster, lower-memory segmentation post-processing

Segmentation mask post-processing was optimized in [PR #24988 by @glenn-jocher](https://github.com/ultralytics/ultralytics/pull/24988), from [@glenn-jocher](https://github.com/glenn-jocher), reducing peak memory usage and improving speed.

Benefits include:

- fewer out-of-memory issues on large images
- better performance in scenes with many objects
- smoother segmentation inference overall

### 🛡 More reliable training checkpoints

A long-standing issue where bad EMA state could block checkpoint saving was fixed in [PR #24986 by @glenn-jocher](https://github.com/ultralytics/ultralytics/pull/24986), from [@glenn-jocher](https://github.com/glenn-jocher).

This means training runs are less likely to fail right at the end because of temporary numerical issues.

* * *

## ⚙ Improvements

### 🧠 Safer mixed-precision attention

Attention computation was reordered in [PR #24990 by @fcakyon](https://github.com/ultralytics/ultralytics/pull/24990), contributed by [@fcakyon](https://github.com/fcakyon), to reduce FP16 overflow risk during AMP and half-precision training.

This helps prevent `inf` and `nan` issues in mixed-precision runs.

### 🧹 Better memory and worker cleanup

Two useful training stability fixes landed in this release:

- [PR #24977 by @glenn-jocher](https://github.com/ultralytics/ultralytics/pull/24977) from [@glenn-jocher](https://github.com/glenn-jocher) fixes `MultiTrainer` memory retention
- [PR #24989 by @fcakyon](https://github.com/ultralytics/ultralytics/pull/24989) from [@fcakyon](https://github.com/fcakyon) properly shuts down stale `DataLoader` workers on reset

Together, these changes help reduce memory buildup and avoid `"too many open files"` errors in long-running jobs.

### 🎯 Improved INT8 calibration for classification exports

Classification export calibration now uses a classification-specific dataset pipeline in [PR #24972 by @onuralpszr](https://github.com/ultralytics/ultralytics/pull/24972), contributed by [@onuralpszr](https://github.com/onuralpszr).

This should produce more correct and robust quantized classification exports.

### 🌍 YOLO-World DDP consistency fix

Distributed training consistency for YOLO-World was improved in [PR #24980 by @Y-T-G](https://github.com/ultralytics/ultralytics/pull/24980), from [@Y-T-G](https://github.com/Y-T-G), by applying evaluation classes on all DDP ranks.

* * *

## ✅ Bug Fixes and Validation Improvements

### 🧾 Better segment dataset validation

Segment tasks now fail early when box-only labels are provided by mistake, thanks to [PR #24992 by @glenn-jocher](https://github.com/ultralytics/ultralytics/pull/24992) from [@glenn-jocher](https://github.com/glenn-jocher).

This should save time by catching invalid dataset formatting earlier.

### 🛠 Safer deconv + batch norm fusion

[PR #24991 by @fcakyon](https://github.com/ultralytics/ultralytics/pull/24991), contributed by [@fcakyon](https://github.com/fcakyon), adds a guard for `fuse_deconv_and_bn` when `bn` is not a `BatchNorm` module.

* * *

## 📚 Documentation and Project Updates

This release also includes a few helpful docs and project improvements:

- [PR #24982 by @oaslananka](https://github.com/ultralytics/ultralytics/pull/24982) from [@oaslananka](https://github.com/oaslananka) adds C++ inference guidance to export docs
- [PR #24984 by @oaslananka](https://github.com/ultralytics/ultralytics/pull/24984) from [@oaslananka](https://github.com/oaslananka) adds tracker state guidance to tracking docs
- [PR #24979 by @onuralpszr](https://github.com/ultralytics/ultralytics/pull/24979) from [@onuralpszr](https://github.com/onuralpszr) adds a Trendshift badge to README files
- [PR #24966 by @glenn-jocher](https://github.com/ultralytics/ultralytics/pull/24966) from [@glenn-jocher](https://github.com/glenn-jocher) linkifies PR numbers in publish Slack notifications

* * *

## 🙌 New Contributor

A big welcome to **[@oaslananka](https://github.com/oaslananka)** for their first contribution in [PR #24982](https://github.com/ultralytics/ultralytics/pull/24982)! 🎉

* * *

## 🧪 Try it out

If you’d like to test the new release locally, you can upgrade with:

```bash
pip install -U ultralytics

```

And for the new LiteRT export flow:

```python
from ultralytics import YOLO

model = YOLO("yolo26n.pt")
model.export(format="litert")

```

You can also explore training and deployment workflows in the **[Ultralytics Platform documentation](https://docs.ultralytics.com/platform)** or use the **[Ultralytics Platform](https://platform.ultralytics.com/)** to annotate datasets, train, and deploy models more easily.

* * *

## 💬 Feedback

We’d love for you to try **v8.4.83** and share your results, especially if you’re working on:

- mobile or embedded deployment
- browser inference
- segmentation workloads
- long-running or multi-dataset training jobs

For the complete set of changes, see the [full changelog from v8.4.82 to v8.4.83](https://github.com/ultralytics/ultralytics/compare/v8.4.82...v8.4.83).

Thanks to everyone in the community who contributed, tested, and shared feedback! 🚀
