# New Release: Ultralytics v8.3.243

**URL:** <https://community.ultralytics.com/t/new-release-ultralytics-v8-3-243/1721>\
**Category:** Discussion\
**Tags:** ultralytics-official, releases, announcements\
**Created:** [December 30, 2025, 1:16pm UTC](https://community.ultralytics.com/t/new-release-ultralytics-v8-3-243/1721 "2025-12-30T13:16:24Z")\
**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:** [December 30, 2025, 1:16pm UTC](https://community.ultralytics.com/t/new-release-ultralytics-v8-3-243/1721/1 "2025-12-30T13:16:25Z")

</div>

# 🚀 Ultralytics v8.3.243 is out! 🧹📡

**TL;DR:** v8.3.243 sharpens training telemetry and console output— **cleaner progress bars** , **richer run metadata** for Ultralytics Platform-style tracking, and **safer/less noisy logging** ✅

If you’re training **Ultralytics YOLO** models like **YOLO11** (recommended), this release is all about making your runs easier to monitor, compare, and debug—especially in notebooks, CI, Docker, and streamed logs.

* * *

## 🌟 New Features (highlights first)

### 🧾✅ Cleaner ConsoleLogger progress bars (dedup + batching)

Training output is now significantly less spammy:

- Suppresses noisy in-progress bar updates (while keeping completion signals)
- Prevents repeated `100%` lines for the same step (epoch/train/val/Class)
- Adds **batched flushing** (by lines/time)
- Safer handler cleanup

Merged in [PR #23082](https://github.com/ultralytics/ultralytics/pull/23082) by [@glenn-jocher](https://github.com/glenn-jocher) and [PR #23079](https://github.com/ultralytics/ultralytics/pull/23079) by [@glenn-jocher](https://github.com/glenn-jocher)

* * *

### 🧠🌍 Richer `training_started` metadata for telemetry

Run tracking gets a big boost with a new `_get_environment_info()` collector, adding:

- OS, Python, hostname
- CPU/GPU details
- command used to launch training
- Git repo/branch/commit (when available)
- model stats (params, GFLOPs, class count)

Included in [PR #23079](https://github.com/ultralytics/ultralytics/pull/23079) by [@glenn-jocher](https://github.com/glenn-jocher)

* * *

### ⏱⬆ Checkpoint upload timing tweak (rate-limit initialization)

The upload timer now initializes at training start, so the **first checkpoint upload waits ~15 minutes** —reducing early-run upload churn.

Included in [PR #23079](https://github.com/ultralytics/ultralytics/pull/23079) by [@glenn-jocher](https://github.com/glenn-jocher)

* * *

## 📚 Docs & Usability Improvements

### 🖥🚫 Headless install guidance promoted (servers/CI/Docker)

Docs now recommend `ultralytics-opencv-headless` for headless environments to avoid OpenCV GUI `libGL` issues.

Shipped in [PR #23077](https://github.com/ultralytics/ultralytics/pull/23077) by [@glenn-jocher](https://github.com/glenn-jocher)

Minimal example:

```bash
pip install ultralytics-opencv-headless

```

* * *

### 🐳⚡ Docker GPU run instructions clarified

Docker examples now include `--runtime=nvidia` to reduce “GPU not found” surprises.

Updated in [PR #23080](https://github.com/ultralytics/ultralytics/pull/23080) by [@lakshanthad](https://github.com/lakshanthad)

* * *

### 🧰🛑 Opt-out of auto requirements checks

New environment variable to skip requirements checks/auto-install behavior (useful for locked/controlled environments):

Added in [PR #23065](https://github.com/ultralytics/ultralytics/pull/23065) by [@onuralpszr](https://github.com/onuralpszr)

```bash
export ULTRALYTICS_SKIP_REQUIREMENTS_CHECKS=1

```

* * *

## 🐛 Small Fixes / Notes

- Pose loss comments corrected in [PR #23060](https://github.com/ultralytics/ultralytics/pull/23060) by [@raimbekovm](https://github.com/raimbekovm)
- YOLO-World docs note a community tip (adding `""` as background class may help in some cases) in [PR #23058](https://github.com/ultralytics/ultralytics/pull/23058) by [@raimbekovm](https://github.com/raimbekovm)
- Dependency update (`zensical>=0.0.15`) in [PR #23052](https://github.com/ultralytics/ultralytics/pull/23052) by [@glenn-jocher](https://github.com/glenn-jocher)

* * *

## ✅ Why this release matters

- **Cleaner logs** 🧹: easier to spot real warnings/metrics in notebooks, CI, and forwarded logs
- **Better reproducibility** 🔍: environment + Git + model metadata improves traceability across machines/containers/branches
- **More predictable syncing** ⏱: reduces early bandwidth spikes from checkpoint uploads
- **Smoother installs** 🧩: fewer dependency headaches on servers and pipelines
- **No major model behavior changes** ✅: focused on logging/telemetry/usability/docs

* * *

## 🔄 Upgrade & try it

Install or upgrade:

```bash
pip install -U ultralytics

```

Then run a quick sanity train with **YOLO11** :

```bash
yolo train model=yolo11n.pt data=coco8.yaml imgsz=640 epochs=1

```

* * *

## 🔎 Release links

You can review the full release details in the [v8.3.243 GitHub Release](https://github.com/ultralytics/ultralytics/releases/tag/v8.3.243), and browse all changes in the [full changelog comparison](https://github.com/ultralytics/ultralytics/compare/v8.3.242...v8.3.243).

* * *

## 💬 Feedback welcome

Try v8.3.243 and let us know:

- Are your training logs noticeably cleaner in your environment?
- Anything you want added to the run metadata payload?
- Any CI/Docker install edge-cases we should cover next?

Thanks to everyone contributing improvements—this kind of polish is what keeps the YOLO ecosystem fast, usable, and reliable.