Our first open-source release. YOLO26-MLX, native YOLO26 on Apple Silicon. No PyTorch. No external GPU. Up to 2.6x faster inference. Up to 1.7x faster training. Accuracy within 0.2% of official results. It won't be the last.

Read the full blog: Check it out on GitHub:

Apple M5 Max YOLO26 Benchmarks - End-to-end predict: 5.95ms (168 FPS) - Forward pass only: 2.8ms (357 FPS)

I wonder how many FPS i can get on M2 Max laptop. This is pretty cool!

Only one way to find out. Let us know what you get!

Will do! Thank you 😊

2.6x faster inference and within 0.2% accuracy by going native MLX instead of PyTorch. The performance was always there in Apple Silicon. The bottleneck was the framework abstraction layer, not the chip.

Nice, gonna give it a try tomorrow Congrats on the launch!

got started with cv through ml5.js because my computer couldn't run anything else. really cool to see native apple silicon inference like this!

Yes! Apple Silicon is removing those barriers so anyone can build, definitely let us know how it goes when you get a chance to run it!

Awesome stuff! Can’t wait to try on my m5 max this weekend

Let us know how it goes!

Thanks for putting MLX on the road.

Thank you for this release, we added support for it in mlx-optiq with near zero loss of accuracy -

So it’s running with Transformers.js or ?

YOLO is a convolution based AI model and not transformers. Implementation is on python using MLX framework.

Ooooo I like that will be checking it out !