MLflow is an open-source platform for keeping track of machine learning and AI work, from logging experiments to versioning models and evaluating LLM apps.
In plain words
Training a model usually means many attempts with different settings. MLflow records each attempt as a run, saving its settings, scores and output files such as model weights. Runs are grouped into experiments so you can compare them. A model registry keeps numbered versions of models and links each one to the run that made it. Newer features cover tracing, evaluating and monitoring LLM and agent applications.
A quick example
Imagine trying ten learning rates for a model. Each training script logs its learning rate and final accuracy to MLflow. Afterwards you open the MLflow interface, sort the ten runs by accuracy, and register the best model as version 1, knowing exactly which run produced it.
Where you'll meet it
- Comparing training runs and their results.
- Versioning models on the way to production.
- Tracing and evaluating LLM and agent apps.
Related
Sources
- DocsMLflow Tracking, MLflow documentation · read 29 Sept 2026
- DocsMLflow Model Registry, MLflow documentation · read 29 Sept 2026
- Docsmlflow/mlflow, MLflow (GitHub) · read 29 Sept 2026
Checked by WhatIsAI against the sources above · 29 Sept 2026. Spotted an error? Report it.