.png)
Weights & Biases
AI developers and machine learning engineers building, training, and deploying models and AI applications.
Overview
Weights & Biases (W&B) is an AI developer platform for training, fine-tuning, and evaluating machine learning models and GenAI applications. Founded in 2017 by Lukas Biewald, Chris Van Pelt, and Shawn Lewis, the company is headquartered in San Francisco and is privately held. W&B provides tools for experiment tracking, hyperparameter optimization, dataset and model versioning, and production monitoring, serving as a system of record for AI development. The platform is used by thousands of companies, including Canva, Microsoft, and Toyota, and has raised a total of $250M in funding.
Funding: Total funding of $250M over 5 rounds, with the first round in May 2018.
What it does
Weights & Biases provides a suite of tools for the machine learning lifecycle. It tracks and visualizes experiments, optimizes hyperparameters with Sweeps, and manages datasets and models with a registry. For GenAI applications, it offers Weave for tracing and debugging, evaluations for rigorous testing, and monitors for production performance. The platform also includes serverless fine-tuning and inference for open-source LLMs, as well as an open-source RL framework (ART) and a reward function automation tool (Ruler).
Key features
- Experiment tracking and visualization
- Hyperparameter optimization (Sweeps)
- Dataset and model versioning (Artifacts, Registry)
- GenAI tracing and debugging (Weave)
- Evaluations with LLM-as-a-judge metrics
- Production monitoring for performance, cost, and health
- Serverless fine-tuning and inference for open-source LLMs
- CoreWeave Sandboxes for isolated agent environments
Use cases
- Training and fine-tuning LLMs
- Developing RAG applications
- Evaluating AI applications for accuracy, latency, and token usage
- Monitoring AI applications in production
- Computer vision model development
- Autonomous driving research
Pricing
| Plan | Price |
|---|---|
| Free | Free |
| Pro | $60 |
| Enterprise | — |
| Personal | Free |
| Advanced Enterprise | — |
Pro plan starts at $60/month billed monthly. Free tier available for personal development and academic research. Enterprise plans require contacting sales. Additional costs for storage and data ingestion beyond included limits.
View current pricingPricing is gathered from the company's public pages and may change. Check the vendor's site before buying.
Pros and cons
Strengths
- Comprehensive platform covering the full ML lifecycle from experimentation to production
- Free tier for personal development and academic research
- Supports a wide range of open-source LLMs for serverless inference
- Used by notable companies like Canva, Microsoft, and Toyota
- Offers both cloud-hosted and privately-hosted deployment options
Limitations
- Pro plan is paid, starting at $60/month, with no free tier for commercial use
- Enterprise features require contacting sales for custom pricing
- Some features like CoreWeave Sandboxes are in public preview and may have limited availability
- Pricing for inference is per model and can vary, with additional costs for data ingestion
What sets it apart
- System of record for AI development with a focus on experiment tracking and reproducibility
- Serverless fine-tuning and inference for a wide range of open-source LLMs
- Open-source RL framework (ART) and automated reward function tool (Ruler)
- Strong emphasis on enterprise security and compliance with HIPAA option and private connectivity
Ecosystem
Notable customers
Frequently asked questions
Is there a free tier?
Yes, W&B offers a Free plan for personal development of AI applications and models, and a Personal plan for self-hosted use. Academic research is free forever with 200GB of cloud storage.
How much does the Pro plan cost?
The Pro plan starts at $60 per month, billed monthly, and includes a 30-day free trial.
Does W&B offer enterprise features?
Yes, the Enterprise plan includes single tenant options, HIPAA compliance, secure private connectivity, customer-managed encryption keys, SSO, and audit logs. Pricing is custom.
Can I run W&B on my own infrastructure?
Yes, the Personal plan allows you to run a W&B server locally on any machine with Docker and Python, and the Advanced Enterprise plan offers flexible deployment options.
What open-source models are supported for inference?
W&B supports serverless inference for models like OpenAI OSS, Meta Llama, DeepSeek, and Z.AI GLM, among others.
Sources
This profile was compiled from the company's own pages and public web research.
Last researched August 11, 2026.
Manage this listing
Own this company? Update your listing information.