Weights & Biases
AI developer platform for tracking ML experiments, versioning models and datasets, and managing workflows.
Updated April 2026
Overview
- Website
- wandb.ai
- Founded
- 2017
- Headquarters
- San Francisco, CA
- Segment
- MLOps & Experiment Tracking
Product overview
Weights & Biases provides W&B Models for training and managing AI models with experiment tracking and a model registry, W&B Weave for evaluating and monitoring AI agents, and Core for reports and automations. Used by over 1,000 companies and 900,000 users including OpenAI, Meta, NVIDIA, Microsoft, Toyota, and Salesforce for productionizing AI at scale. Distinct for flexible SaaS, dedicated, or local deployment options, enterprise compliance (SOC 2, HIPAA, ISO), and seamless integrations across the ML lifecycle.
Revenue model
Freemium: Free tier ($0, 5GB storage, 5 seats); Teams $50/user/month plus usage ($0.03/GB over 100GB, $1/hour over 5,000 tracked hours/year); Enterprise custom pricing with unlimited features, SSO, audit logs, dedicated support, and flexible deployment. Free for academics (100GB storage, 100 seats).
Moat
Weights & Biases' key competitive moat is its high switching costs from deep integration into ML workflows, where teams log experiments, track lineage, version models, and collaborate across thousands of runs, making migration disruptive as seen in customer dependencies at Square, RBC, and IBM. This is reinforced by proprietary tools like Weave for agent evaluation and serverless fine-tuning, plus network effects from a shared registry of datasets/models/prompts among leading AI teams.