Weights & Biases

Weights & Biases (W&B) is an MLOps and LLMOps platform that gives ML and AI teams experiment tracking, model and dataset versioning, hyperparameter optimization, and a system of record for training models and developing GenAI applications. It pairs W&B Models for the ML lifecycle with W&B Weave for tracing and evaluating LLM and agent applications.

Weights & Biases is a developer-first AI platform that helps machine learning and AI teams track, visualize, and reproduce their work across the full model lifecycle. Its W&B Models product provides experiment tracking, hyperparameter sweeps, a model and artifact registry, and collaborative reports for teams training and fine-tuning models, while W&B Weave delivers tracing, evaluations, guardrails, and production monitoring for teams building LLM and agent applications. The platform integrates natively with frameworks like PyTorch, TensorFlow, Keras, and Hugging Face Transformers, and is offered as a managed SaaS cloud, a dedicated cloud, or self-hosted server deployment. Following CoreWeave's acquisition of the company (announced March 2025, completed May 2025), W&B has expanded into serverless inference, serverless training/RL, and sandboxes running on CoreWeave infrastructure. It is used by leading AI labs and enterprises including OpenAI, Meta, NVIDIA, Cohere, Toyota, Salesforce, and Microsoft.

Category

Developer Tools

Pricing

Freemium, Subscription, Usage-based, Contact for pricing

Target Market

Data Scientists, ML Engineers, AI Researchers, Enterprise Developers, CTOs

Founded

2017

Headquarters

San Francisco, California, USA

Customers

1,000+ companies and over 1 million AI engineers/practitioners

• ✓
  Experiment Tracking

  Logs metrics, hyperparameters, code, model weights, and sample predictions so training runs can be visualized, compared, and reproduced.

• ✓
  Sweeps (Hyperparameter Optimization)

  Automates hyperparameter search across runs to find optimal model configurations.

• ✓
  Model & Artifact Registry

  Versions and manages datasets, models, and ML pipelines with lineage tracking for governance and reproducibility.

• ✓
  W&B Weave

  Provides tracing, evaluations, guardrails, and production monitoring for LLM and agent applications.

• ✓
  Reports

  Creates interactive, shareable dashboards to document and collaborate on ML insights across a team.

• ✓
  Serverless Inference & Training

  Offers hosted open-source LLMs plus serverless fine-tuning with RL and SFT, powered by CoreWeave infrastructure.

• •
  Foundation model training

  AI labs use W&B Models to track large-scale pre-training and fine-tuning runs and debug infrastructure issues.

• •
  GenAI application development

  Teams use W&B Weave to trace, evaluate, and monitor LLM and agent applications from experimentation to production.

• •
  Hyperparameter tuning at scale

  Data scientists run Sweeps to systematically optimize model performance across many configurations.

Free tier and a self-hosted Personal tier are $0; Pro starts around $60/mo with usage-based overages for storage ($0.03/GB) and Weave data ingestion ($0.10/MB). Enterprise and Advanced (self-hosted) tiers are custom-priced. Free Pro-tier access is available for qualifying academics.

This page was written from 9 sources, 7 on domains other than wandb.ai.

1. 1.wandb.ai — about usvendor
2. 2.wandb.ai — pricingvendor
3. 3.insightpartners.com — weights and biases raises 45m series b to expand beyond expe
4. 4.research.contrary.com — weights biases
5. 5.prnewswire.com — coreweave completes acquisition of weights biases 302445966
6. 6.coreweave.com — coreweave to acquire core scientific
7. 7.docs.wandb.ai — get started
8. 8.docs.wandb.ai — integrations
9. 9.zenml.io — wandb pricing