Nyckel

Nyckel is a machine-learning platform for building custom prediction models without in-house ML expertise. Developers upload labeled examples of images, text or structured data, and Nyckel automatically tests many model types to reach a target accuracy, then hosts the chosen model behind an API. Supported tasks include classification, object detection and segmentation, and the platform offers pretrained classifiers and detectors, active learning suggestions to improve accuracy, human review queues and decision thresholds for automation.

Nyckel positions itself for repetitive decisions such as spam or fraud detection, document classification, content moderation, product categorization and defect detection, where consistent outputs and confidence scores matter more than generated text. Model quality depends on the examples provided and was not tested for this listing.

The free plan costs $0 per month and includes 100 invokes, up to 200 samples and five functions, with additional invokes at $0.005 each. Starter costs $149 per month for 15,000 invokes, up to 30,000 samples and custom-trained models, and Business costs $599 per month for 300,000 invokes, up to 600k samples and automation features such as multi-model pipelines and prediction auditing. Enterprise plans with dedicated or regional infrastructure are priced individually.

Nyckel's terms are an order form under the Common Paper Cloud Service Agreement standard terms version 2.0. The company states that customer data and models are not shared with other customers and that it is SOC 2 certified and HIPAA compliant; these are vendor statements that were not independently verified.