Open Source Alternatives to Weights & Biases
A machine-learning platform that tracks experiments, visualizes results, manages model metadata, supports collaborative team work, integrates with popular data science tools, and provides cloud storage for model artifacts.
Whether you want to cut software costs or escape vendor lock-in, these open source alternatives to Weights & Biases give you an option you fully own. Run them on your own server or desktop, so your data stays under your control. This page lists 6 open source alternatives to Weights & Biases. The most popular are Opik, Transformer Lab and Latitude. Most use the Apache-2.0 or MIT license, and 6 offer an official Docker image.
Opik
An observability and evaluation platform for monitoring, testing, and optimizing generative AI and agentic systems.
Transformer Lab
A unified environment to train, evaluate, and scale large language and diffusion models on local or cluster hardware.
Latitude
Monitors AI agents in production by clustering failures into issues and generating evaluations from real traces.
Agenta
An LLMOps platform for prompt management, evaluation, and observability to help teams build reliable AI applications.
OpenLIT
Provides observability, distributed tracing, and prompt management for generative AI and large language model applications.
mlop
A self hosted framework for machine learning engineers to track metrics, parameters, and gradients during model training.
Join our newsletter to get shiny new open source software delivered to your inbox. Unsubscribe anytime.