# Model Evaluation Once your dataset contains predictions, FiftyOne’s [evaluation API](../user_guide/evaluation/index.md#evaluating-models) computes standard metrics (mAP, precision/recall, confusion matrices) and — more importantly — lets you drill into the individual successes and failures behind them. Use the [Model Evaluation panel](../user_guide/app.md#app-model-evaluation-panel) to explore and compare evaluation runs in the App, [scenario analysis](../user_guide/evaluation/index.md#analyzing-scenarios) to compare performance across subsets of your data, and [mistakenness](../brain/index.md#brain-label-mistakes) to separate model errors from label errors.