# New Release: Ultralytics v8.4.35

**URL:** <https://community.ultralytics.com/t/new-release-ultralytics-v8-4-35/1926>\
**Category:** Discussion\
**Tags:** ultralytics-official, releases, announcements\
**Created:** [April 8, 2026, 1:34pm UTC](https://community.ultralytics.com/t/new-release-ultralytics-v8-4-35/1926 "2026-04-08T13:34:06Z")\
**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 8, 2026, 1:34pm UTC](https://community.ultralytics.com/t/new-release-ultralytics-v8-4-35/1926/1 "2026-04-08T13:34:06Z")

</div>

# Ultralytics v8.4.35 is out 🎉

**Quick summary:** Ultralytics `v8.4.35` is a **stability-first release** focused on smarter training recovery, safer dataset caching, cleaner runtime behavior, and better compatibility across training and inference workflows. If you run long trainings, debug custom datasets, or deploy on edge/OpenVINO setups, this is a very worthwhile upgrade 🚀

You can explore the full release on [GitHub Releases](https://github.com/ultralytics/ultralytics/releases/tag/v8.4.35) and review every code change in the [full changelog](https://github.com/ultralytics/ultralytics/compare/v8.4.34...v8.4.35).

* * *

## 🌟 Highlights

### 🔁 Smarter NaN recovery during training

The biggest change in `v8.4.35` comes from [PR #24154 by @glenn-jocher](https://github.com/ultralytics/ultralytics/pull/24154) with contributions from [Glenn Jocher](https://github.com/glenn-jocher).

- Training now recovers from `last_good.pt` instead of retrying a potentially corrupted `last.pt`
- Checkpoint saving is skipped when EMA weights contain `NaN` or `Inf`
- This helps prevent bad checkpoints from propagating through retries

**Why it matters:** fewer failed long runs, less wasted GPU time, and more trustworthy resume behavior 🛡

### 🗂 Safer and clearer detection dataset caching

Also in [PR #24154 by @glenn-jocher](https://github.com/ultralytics/ultralytics/pull/24154):

- Empty detection caches are no longer written
- Corrupt label failures now report clearer final errors
- Detection dataset checks now accept a dataset directory directly, not just a YAML file
- NDJSON-to-YOLO cached conversions are rebuilt when expected split folders are missing

**Why it matters:** debugging data issues is now faster and much less frustrating 🧪

### ⚙ More robust OpenVINO inference

[PR #24156 by @glenn-jocher](https://github.com/ultralytics/ultralytics/pull/24156) from [@glenn-jocher](https://github.com/glenn-jocher) improves OpenVINO runtime behavior by:

- Making device fallback logic safer
- Forcing FP32 inference on Linux ARM64 CPU for better compatibility and stability

**Why it matters:** more predictable deployment behavior, especially on edge and ARM systems 🏭

* * *

## Improvements

### 📈 Better experiment resume behavior

[PR #24110 by @artest08](https://github.com/ultralytics/ultralytics/pull/24110) from [@artest08](https://github.com/artest08) updates W&B resume behavior so resumed training continues the original run instead of creating a disconnected new one.

### 🧠 Better training config handling

[PR #23640 by @artest08](https://github.com/ultralytics/ultralytics/pull/23640) from [@artest08](https://github.com/artest08) prevents duplicate pretrained weight loading in `Model.train()` for YAML-based model configs.

### 🔕 Cleaner logs

[PR #24154 by @glenn-jocher](https://github.com/ultralytics/ultralytics/pull/24154) throttles repeated `Platform: Model not found` warnings to reduce noisy logs during runs.

### 🔧 Broader compatibility updates

- [PR #24152 by @Y-T-G](https://github.com/ultralytics/ultralytics/pull/24152) from [@Y-T-G](https://github.com/Y-T-G) adds backward compatibility for older YOLO12 Area Attention checkpoints
- [PR #24143 by @fcakyon](https://github.com/ultralytics/ultralytics/pull/24143) from [@fcakyon](https://github.com/fcakyon) adds missing `isatty()` support in `_ConsoleCapture`, preventing crashes with some libraries such as `transformers`
- [PR #24145 by @ausk](https://github.com/ultralytics/ultralytics/pull/24145) from [@ausk](https://github.com/ausk) fixes a SAM3 decoder cache-miss dtype issue affecting coordinate retrieval

* * *

## Docs and ecosystem updates 📚

This release also includes several documentation improvements to better reflect the current Ultralytics YOLO ecosystem, with [Ultralytics YOLO26](https://platform.ultralytics.com/ultralytics/yolo26) continuing as the recommended model line for new projects.

- [PR #24140 by @raimbekovm](https://github.com/ultralytics/ultralytics/pull/24140) from [@raimbekovm](https://github.com/raimbekovm) refreshes FastSAM and MobileSAM docs with YOLO26 benchmark context and clearer reproducibility notes
- [PR #24133 by @RizwanMunawar](https://github.com/ultralytics/ultralytics/pull/24133) from [@RizwanMunawar](https://github.com/RizwanMunawar) updates SAHI tiled inference docs for YOLO26
- [PR #24139 by @raimbekovm](https://github.com/ultralytics/ultralytics/pull/24139) from [@raimbekovm](https://github.com/raimbekovm) modernizes outdated model references toward YOLO26
- [Ultralytics Platform](https://platform.ultralytics.com) remains the easiest place to annotate datasets, train, deploy, and monitor YOLO workflows end to end

* * *

## Bug fixes

A few additional fixes landed in this release as well:

- [PR #24146 by @Laughing-q](https://github.com/ultralytics/ultralytics/pull/24146) from [@Laughing-q](https://github.com/Laughing-q) fixes outdated `YAML.dump` usage in `benchmark.py`
- [PR #24127 by @ahmet-f-gumustas](https://github.com/ultralytics/ultralytics/pull/24127) from [@ahmet-f-gumustas](https://github.com/ahmet-f-gumustas) adds missing type hints to `segment2box` and `scale_boxes` in `utils/ops.py`

* * *

## New contributor 🙌

A big welcome to [@ausk](https://github.com/ausk), who made their first contribution in [PR #24145](https://github.com/ultralytics/ultralytics/pull/24145)! Thanks for helping improve the project 💙

* * *

## Why upgrade?

`v8.4.35` is a **low-risk, high-value** update that improves reliability without changing your overall workflow:

- More trustworthy recovery when training encounters NaNs
- Better dataset validation and clearer cache behavior
- Cleaner logs during long runs
- More stable OpenVINO inference on ARM and fallback scenarios
- Smoother experiment tracking resumes

If you’re training frequently, validating custom datasets, or deploying production pipelines, this release should make day-to-day work noticeably smoother ✅

* * *

## Try it out

Update with:

```bash
pip install -U ultralytics

```

Then let us know how `v8.4.35` works in your pipelines, especially around resume behavior, dataset validation, and OpenVINO deployment. Feedback, edge cases, and regression reports are always appreciated 🙏