Bittensor subnet
Perturb
Tests your image AI model with tricky inputs and sends a report on where it breaks.
τ
ב
Can I use it today?
Open Perturb ↗Yes, you can use it today.
Web app. Pricing not listed.
TAO in its pool1.7K#108 of 128 subnets
Share of new TAOUnder 0.1%Next to nothing right now
Its tokenב0.0026 TAO each
In this slot sinceMay 2026159 days running
What you can do with it
- Order a $99 robustness scan of an image classification model
- Try a live demo that makes an image designed to fool AI
- Contact them for enterprise testing with CI/CD and SLAs
Things to know
Only image classification models are supported now, the Pro plan is early access and orders go through a contact form.
TAO in its pool, month by month
Guides that recommend it
Similar subnets
SN34BitMindSpots deepfakes and AI-generated content so you can check it before you trust or share it.SN37AureliusProba predicts what a training dataset will teach an AI model before you pay to train it.SN56GradientsTrain your own custom AI image or text model without any machine learning know-how.SN9iotaTrain your own AI models across many scattered machines, either yourself with an SDK or managed.
Sources
Checked October 13, 2026. Pool and emission numbers read from the chain October 13, 2026.