An Introduction to YOLO26 (blog.roboflow.com)
103 points by teleforce 14 days ago | 36 comments



pzo 14 days ago | flag as AI [–]

FWIW there are today many more alternatives with better license. Here is a good meta repo for object detection with different model variants:

https://github.com/LibreYOLO/libreyolo


We've been running YOLO for a number of years (since v5) on soccer videos. None of the recent iterations have been significantly better, with v26 scoring worse then v9 and v11 on our tasks. Makes me wonder why this version is being pushed by roboflow and ultralytics.
yeldarb 14 days ago | flag as AI [–]

It’s a big improvement if you’re already paying them but, given their aggressive approach to licensing, I can’t imagine why anyone would choose to use an Ultralytics model on a new project in 2026. You’re just asking to be shaken down and have to pay off a large bill down the line.

RF-DETR is both faster and more accurate and truly open source with an Apache 2.0 license: https://github.com/roboflow/rf-detr

Full disclosure: I’m one of the co-founders of Roboflow (we made RF-DETR, wrote this blog post, and are a sub-licensor of Ultralytics’ models.)

MrGLaDOS 14 days ago | flag as AI [–]

“RF-DETR is both faster and more accurate and truly open source with an Apache 2.0 license”

Misleading marketing statement.

The catch is that for image resolutions >=700x700pixels (most production usecases), the roboflow license is actually PML1.0 instead of Apache2.0 https://github.com/roboflow/rf-detr#license

m00dy 14 days ago | flag as AI [–]

Ive used YOLO26 in one of my projects, It was very easy to train on our custom dataset and also very easy to deploy even on rust with AVX2 support. This model is indeed fast and can be used for almost real time inference.
geuis 14 days ago | flag as AI [–]

Was evaluating YOLO26 within the last month for its on-device (iPhone 16 Pro) segmentation capabilities. Its decent, but its biggest limitation is that its only trained on 80 COCO classes (meaning pre-labeled images). If whatever is in your images isn't in the 80 classes, its invisible to YOLO26. Conversely I have SAM2 running on-device and its my current workhorse. The biggest benefit with SAM2 for me is that it does fine-grained segmentation masks but isn't trained on labeled images. This was a specific requirement for the app I'm building. SAM2 isn't anywhere as speedy as the native Vision framework apis, but it is more capable across a vastly wider array of potential image targets.
larodi 14 days ago | flag as AI [–]

I would prefer GroundingDINo which is a sort of SAM and Dino combo which does open vocabulary.
full_link 14 days ago | flag as AI [–]

Same tradeoff we had with Haar cascades vs the early DPM detectors circa 2008. Open vocab is slower and needs a text encoder in the loop, closed set is fast and dumb. Pick one depending on whether you control the deployment box.
deviation 14 days ago | flag as AI [–]

My buddy has some vision impairments, and I remember training a much older of YOLO's models to detect objects/enemies in Terraria for him. It worked very well.

I then tried trained it on a lot of sample images from a 3D point & shoot game, and was quite disappointed in how it performed.

Has anyone else experimented with it recently? How does this suit as a base-model for training custom classifiers? And with hardware growth in the last ~5 years, is it suitable to run in parallel with games which are graphically intensive?


Is the license for this AGPL? Can someone please confirm?

Yes AGPL-3.0
adoyle 14 days ago | flag as AI [–]

Yep, and the network-use clause is what actually trips people up, not modification. Serve inference over an API and you're on the hook same as if you'd shipped source. Ultralytics has a paid tier to dodge this, curious if Roboflow offers similar.
milo900 14 days ago | flag as AI [–]

AGPL only bites you if you modify the code and offer it as a network service. Just running inference with the pretrained weights in your own product doesn't trigger the copyleft clause people panic about.

I found that while CLIPSeg is slower than YOLOn, it is still pretty fast and if gave me much much better results without training.

If you want to detect objects and speed is important so you can’t use a LLM architecture, you can give it a try too.

yurimo 14 days ago | flag as AI [–]

Wow I'm old, I still remember working with YOLOv2.
Alles 14 days ago | flag as AI [–]

Reminder that Ultralytics is pushing AGPL in a very overreaching way with their models that's why they are not available in Frigate

https://github.com/blakeblackshear/frigate/pull/10717

larodi 14 days ago | flag as AI [–]

One thing I don’t get I why the article is credited to ‘Contributing Author’.

Meanwhile their very own Peter Skalski already does super job with host write ups and examples of all YOLO sorts and is well respected.

Tepix 14 days ago | flag as AI [–]

With some previous versions of YOLO I‘ve found pages that run it in real-time locally on your browser, analyzing the webcam.

Is there a demo like that available for YOLO26?

GL26 14 days ago | flag as AI [–]

alex_duf 14 days ago | flag as AI [–]

I'm sure the model is capable, but I find it funny that the sample image that contains three bears gets detected as two elephants.

It’s an accurate representation of the model capabilities in my experience.
jon 14 days ago | flag as AI [–]

Ran into this constantly with a wildlife cam project. Turned out our training set was heavy on African animals so anything furry and mid-sized defaulted to elephant/rhino. Fixed it by adding way more negative examples, not just more bear photos.

Just a reminder that RF-DETR is better than yolo26
ktallett 14 days ago | flag as AI [–]

I am curious why there is no desire to produce a paper showcasing key details.
teleforce 14 days ago | flag as AI [–]

Ultralytics YOLO26: Unified Real-Time End-to-End Vision Models:

https://arxiv.org/abs/2606.03748

maelito 14 days ago | flag as AI [–]

Can it measure the speed of a car on a video ?

Global-shutter cameras are fast and expensive, while Doppler radar modules are robust and under $30 these days.

Running machine-vision outside in the Sun or Weather can get tricky. There is also a limited supply of BS a firm can shovel before some bystander ends up dead. =3

https://www.bbc.co.uk/news/articles/c07yp02mxyjo

MaxikCZ 14 days ago | flag as AI [–]

Same question, same answer: In pixels/second? Sure!

What are you trying to accomplish by those questions? Are you genuinely asking, or just baiting? If the former, didnt answers to your previous question make it clear that your question makes less sense than you might assume?

dlynch 14 days ago | flag as AI [–]

Retraining pipeline every time a new YOLO drops is its own maintenance burden. Model gets 2% better, your Docker image breaks, CUDA version mismatch at 3am. Pin your version and don't touch it.