Pricing

A function is a single AI decision exposed through a stable API endpoint. Each function answers one question for your application — for example "what product is in this image?" or "is this message spam?". Nyckel abstracts away the underlying machine learning complexity and automatically evaluates and selects the best model as real data flows through.

Models are the underlying AI systems that produce predictions. Nyckel automatically evaluates generalist models (GPT, Claude, Gemini) and foundation models optimised for visual and structured tasks (CLIP, DINO for classification, detection, and segmentation). The best-performing model for each function is selected and updated automatically over time — you don't need to choose or manage models yourself.

An invoke is an inference request. It occurs when your application sends data — an image, document, or piece of text — to a function and asks the model for a prediction. The response includes both the predicted decision and a confidence score between 1 and 100 indicating how certain the model is.

Samples are data points captured from real production usage. When a function is invoked, Nyckel stores the input along with the prediction, confidence score, and an embedding — a machine learning representation of the data. These embeddings power automatic training, outlier detection, surfacing of difficult examples for review, and efficient AutoML processes that improve model accuracy over time.

The Free plan gives you access to AI models that are ready instantly — great for testing and early development. Starter adds custom-trained models that improve automatically from real usage and feedback, private inference (your data is not sent to public AI models), and smart review suggestions that surface rare or confused predictions for you to review. Upgrade to Starter when you need higher accuracy or when your data is sensitive.

Each plan includes a set number of invokes and samples per month at no extra charge. If you exceed those limits, additional usage is billed at the per-unit rate shown on your plan. Invokes are counted each time your application calls a function for a prediction. Samples accrue as predictions are captured and stored for review and training.

Yes. You can upgrade or downgrade your plan at any time from your account settings. Upgrades take effect immediately. If you have questions about which plan is right for you, contact us and we're happy to help.

Additional invokes and samples are billed at the overage rates shown on your plan card above. There are no surprise charges — usage is metered transparently and you can monitor consumption in your dashboard at any time.