Pricing that works for everyone

We define ingested bytes as bytes that we receive, process, and store on your behalf. This includes trace metadata, LLM inputs/outputs, and any other information you explicitly log to Weave, but does not include communication overhead (e.g., HTTP headers) or any other data that is not placed in long term storage. We count bytes as "ingested" only once at the time they are received and stored.

We recommend using the free trial to estimate how much data ingestion you can expect to use. An average trace is approximately 0.22 MB, but the distribution is wide and varies depending on the use case and type of data sent.

Each inference API has different costs for input and output tokens. Explore detailed pricing.

A tracked hour is wall-clock time when training a model. If your model takes 8 hours to train, you have used 8 tracked hours. Today, this is irrespective of GPU quality, speed, etc. Importing runs from other platforms or syncing offline runs will contribute to your organization's tracked hour limit.

Storage includes both artifacts and data logged to runs.

Weights & Biases calculates your storage usage over the last 30 days. For example, if you use 100 GB of storage for 15 days of March, 200 GB for 1 day of March, and 300GB over 14 days of March your storage usage would be:

100 GB x 15 days = 1500 GB-days
200 GB x 1 day x = 200 GB-days
300 GB x 14 days x = 4200 GB-days
5900 GB-days / (30 days) = 196.6 GB

At the end of 30 days, Weights & Biases rounds your storage to the nearest MB. Therefore, your storage usage for March would be 197 GB.

Weights & Biases provides a free Pro license to academic institutions pursuing research not connected to a for-profit entity. This license is intended for students, professors, and postdoctoral researchers. An active email address affiliated with an academic institution is required.