What it is
Weave is the LLM-focused layer on top of Weights & Biases — the same platform many teams already use for ML experiment tracking, model registry, and dataset versioning. Trace LLM calls with the @weave.op decorator, run evaluations, browse leaderboards, and manage prompts alongside the rest of your ML workflow.
Free tier: limited seats, storage, ingestion. Paid from $60/month.
Where it shines
- Continuity with W&B. If your ML org has been on W&B for years, Weave is the path of least resistance for adding LLM observability.
- Decorator instrumentation.
@weave.opis a clean API — wrap a function, get a trace. - Leaderboards. The eval-comparison UI inherits W&B's strong charting heritage.
Where it falls short
- Standalone product. Weave alone, without W&B context, isn't the strongest pick in any category.
- PM collaboration. Versioning and review workflows trail the eval-first specialists.
- LLM-first orientation. It's still an MLOps platform with LLM features, not the other way around.
Bottom line
The right pick if you're already a W&B shop and your LLM work needs to live in the same neighborhood as your ML pipelines. Outside that audience, the LLM specialists (Braintrust, Langfuse) deliver more value per dollar.