Weights & Biases

Weights & Biases (W&B) is an AI developer platform used by machine learning teams to track experiments, manage models, and evaluate and monitor GenAI applications. The platform covers four product surfaces: W&B Models (experiment tracking, hyperparameter sweeps, artifact registry), W&B Weave (LLM observability, tracing, and evaluations), W&B Inference (serverless access to 20+ open-weight models), and CoreWeave Sandboxes for GPU-backed training. Originally founded in 2018, W&B became the de-facto MLOps standard with over one million users including OpenAI, Meta, NVIDIA, and Toyota, and was acquired by CoreWeave in 2025 for a reported ~$1.7 billion.