# New Release: Ultralytics v8.4.34

**URL:** <https://community.ultralytics.com/t/new-release-ultralytics-v8-4-34/1922>\
**Category:** Discussion\
**Tags:** ultralytics-official, releases, announcements\
**Created:** [April 7, 2026, 1:33pm UTC](https://community.ultralytics.com/t/new-release-ultralytics-v8-4-34/1922 "2026-04-07T13:33:19Z")\
**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:** [April 7, 2026, 1:33pm UTC](https://community.ultralytics.com/t/new-release-ultralytics-v8-4-34/1922/1 "2026-04-07T13:33:19Z")

</div>

# Ultralytics v8.4.34 is out 🚀

**Quick summary:** Ultralytics `v8.4.34` is a tuning and stability-focused release with one standout new capability: **multi-dataset hyperparameter tuning in a single run** 🧠. This update also brings more reliable training resume behavior, thread-safe ONNX export, several core runtime fixes, and broad documentation refreshes across **Ultralytics YOLO** and the [Ultralytics Platform](https://platform.ultralytics.com).

If you’re training across mixed domains, exporting in parallel workflows, or deploying on edge devices, this release is definitely worth a look 👇

## 🌟 Highlights

### 🧠 Multi-dataset hyperparameter tuning

The biggest addition in `v8.4.34` comes from [PR #24067 by @Laughing-q](https://github.com/ultralytics/ultralytics/pull/24067) with support for passing **multiple datasets** to `model.tune()`.

What’s new:

- `data` can now be a single dataset or a list of datasets
- Each tuning iteration trains across all provided datasets
- Results are combined and **fitness is averaged across datasets**
- Tuning decisions now reflect overall performance instead of overfitting to just one dataset

This is especially useful for teams training one model across different domains like:

- color + grayscale data
- multiple cameras or environments
- blended internal/external datasets

A minimal example looks like this:

```python
from ultralytics import YOLO

model = YOLO("yolo26n.pt")
model.tune(data=["coco8.yaml", "coco8-grayscale.yaml"], epochs=5, iterations=10)

```

## ✅ Stability and reliability improvements

### 🛡 Safer training resume on small datasets

[PR #24085 by @Y-T-G](https://github.com/ultralytics/ultralytics/pull/24085) fixes a resume-related loss spike issue by keeping AdamW’s `exp_avg_sq` state in FP32. This helps reduce instability when continuing training from checkpoints, especially on smaller datasets.

### 🔒 Thread-safe ONNX export

[PR #24092 by @glenn-jocher](https://github.com/ultralytics/ultralytics/pull/24092) adds export locking to prevent collisions in PyTorch’s global ONNX exporter state when exports run concurrently. A regression test was also included for parallel export safety.

### ⚙ Core runtime robustness

Several fixes improve reliability across training and inference workflows:

- [PR #24056 by @nameearly](https://github.com/ultralytics/ultralytics/pull/24056) prevents DDP cleanup issues when command generation fails
- [PR #24114 by @ZoomZoneZero](https://github.com/ultralytics/ultralytics/pull/24114) fixes `AAttn` crashes when `dim` is not divisible by `num_heads`
- [PR #24115 by @Y-T-G](https://github.com/ultralytics/ultralytics/pull/24115) clamps negative coordinates in `crop_mask()` for safer segmentation postprocessing
- [PR #24099 by @onuralpszr](https://github.com/ultralytics/ultralytics/pull/24099) updates `draw_specific_kpts()` to respect user-specified keypoint index order and safely handle missing confidence values

## 📚 Docs and ecosystem updates

This release also includes a broad docs refresh with stronger **YOLO26** coverage, updated edge guidance, and Platform improvements.

### Edge and deployment docs

- [PR #24068 by @lakshanthad](https://github.com/ultralytics/ultralytics/pull/24068) updates NVIDIA Jetson AGX Orin benchmarks with YOLO26
- [PR #24118 by @lakshanthad](https://github.com/ultralytics/ultralytics/pull/24118) updates NVIDIA Jetson Orin NX 16GB benchmarks with YOLO26
- [PR #24097 by @lakshanthad](https://github.com/ultralytics/ultralytics/pull/24097) updates NVIDIA Jetson Orin Nano Super benchmarks with YOLO26
- [PR #24081 by @lakshanthad](https://github.com/ultralytics/ultralytics/pull/24081) adds missing `cuDSS` guidance for JetPack 6 installs
- [PR #24141 by @onuralpszr](https://github.com/ultralytics/ultralytics/pull/24141) updates the DeepStream 7.1 docs link for JetPack 6.1
- [PR #24142 by @onuralpszr](https://github.com/ultralytics/ultralytics/pull/24142) expands DeepStream install instructions for JetPack 7.1 with DeepStream 9.0

### SAM documentation refresh

- [PR #24071 by @raimbekovm](https://github.com/ultralytics/ultralytics/pull/24071) updates SAM docs with YOLO26 references and refreshed benchmarks
- [PR #24072 by @raimbekovm](https://github.com/ultralytics/ultralytics/pull/24072) updates SAM 2 docs with YOLO26 references and refreshed benchmarks
- [PR #24045 by @raimbekovm](https://github.com/ultralytics/ultralytics/pull/24045) updates SAM 3 docs with YOLO26 references and corrected model sizes

### Ultralytics Platform improvements

- [PR #24037 by @laodouya](https://github.com/ultralytics/ultralytics/pull/24037) improves documentation for YOLO models in Platform Smart Annotation
- [PR #24087 by @sergiuwaxmann](https://github.com/ultralytics/ultralytics/pull/24087) adds docs for split redistribution
- [PR #24060 by @yogendrasinghx](https://github.com/ultralytics/ultralytics/pull/24060) improves Platform Settings docs with clearer tabs and screenshots
- [PR #24090 by @sergiuwaxmann](https://github.com/ultralytics/ultralytics/pull/24090) updates the docs banner
- [PR #24091 by @sergiuwaxmann](https://github.com/ultralytics/ultralytics/pull/24091) fixes the banner link

### Additional maintenance and docs updates

- [PR #24082 by @dependabot[bot]](https://github.com/ultralytics/ultralytics/pull/24082) bumps `onnx` to `1.21.0rc1` in the RT-DETR ONNX Runtime Python example
- [PR #24088 by @Laughing-q](https://github.com/ultralytics/ultralytics/pull/24088) removes a redundant Ray Tune test
- [PR #24094 by @dependabot[bot]](https://github.com/ultralytics/ultralytics/pull/24094) bumps `onnx` to `1.21.0`
- [PR #24098 by @RizwanMunawar](https://github.com/ultralytics/ultralytics/pull/24098) adds a docs video resource
- [PR #24101 by @onuralpszr](https://github.com/ultralytics/ultralytics/pull/24101) updates the Snyk badge

## 🎯 Why this release matters

This update improves three important parts of the workflow:

- **Better tuning quality** : multi-dataset tuning helps produce models that generalize better across varied data sources
- **More dependable pipelines** : resume training and ONNX export are now safer in real-world production setups
- **Better deployment guidance** : refreshed YOLO26 and Jetson docs make edge decisions easier and more current

For new projects, **YOLO26** remains the recommended model family, with smaller, faster, and more accurate models than YOLO11 across all supported tasks.

## 🙌 New contributors

A big welcome to these first-time contributors 🎉

- [@yogendrasinghx in PR #24060](https://github.com/ultralytics/ultralytics/pull/24060)
- [@ZoomZoneZero in PR #24114](https://github.com/ultralytics/ultralytics/pull/24114)
- [@nameearly in PR #24056](https://github.com/ultralytics/ultralytics/pull/24056)
- [@laodouya in PR #24037](https://github.com/ultralytics/ultralytics/pull/24037)

## 🔗 Try it out

You can upgrade with:

```bash
pip install -U ultralytics

```

Then explore the full release in the [v8.4.34 release page](https://github.com/ultralytics/ultralytics/releases/tag/v8.4.34) or browse every change in the [full changelog from v8.4.33 to v8.4.34](https://github.com/ultralytics/ultralytics/compare/v8.4.33...v8.4.34).

If you give `v8.4.34` a try, we’d love to hear how it performs for your training, export, and deployment workflows 💬