Skip to content
effecthub
Verified Oct 15, 2026

Weights & Biases

Verified Oct 15, 2026 Discontinued

Folded into CoreWeave Forge on 30 September 2026; wandb.ai redirects to a Forge login Other discontinued tools

Use tool
Share

Experiment tracking, model registry and LLM evaluation, now sold as part of CoreWeave Forge

experiment trackingmlopsllm evaluationmodel monitoringFreemium · from $60/mo

Listed on EffectHub ·

Web · API · iOS
Screenshot of the Weights & Biases website
#747
wandb.ai as we captured it — our own screenshot, not the vendor’s press image.

Pricing

Pricing model
Freemium
Paid options from
$60/mo
Billing
Monthly

Free: up to 5 model seats, 5 GB storage and 1 GB of Weave ingestion a month. Pro: $60 a month minimum, billed monthly, for teams under 50 employees, with up to 10 seats, 100 GB storage and 1.5 GB Weave ingestion, overages at $0.03 per GB storage and $0.10 per MB of Weave data, 30-day trial. Enterprise is quoted (Weave overage $0.20 per MB); an Academic plan is free with 100 seats and 200 GB. Inference is billed separately.

Use tool ↗
Overview

About Weights & Biases

Weights & Biases logs training runs, metrics, hyperparameters and artifacts so experiments can be compared and reproduced, keeps datasets, models and prompts in a registry, and traces and evaluates LLM applications through Weave. CoreWeave bought the company in May 2025 and on 30 September 2026 folded it into CoreWeave Forge, a bundle that adds serverless RL training, hosted inference, sandboxes and notebooks around the W&B tools; SaaS accounts now sign in through Forge with their existing credentials, while self-hosted and dedicated instances are unchanged. The SDK, an iOS app for watching runs, and an MCP server for agents remain. Pricing moved to the CoreWeave site: a free tier for small teams, Pro from $60 a month, and enterprise contracts.

Platforms
Web · API · iOS
API
Yes
Last verified
Oct 15, 2026

Verified sources

  • CoreWeave Forge pricing: Free $0 with up to 5 seats, 5 GB storage and 1 GB Weave ingestion; Pro $60/month minimum with up to 10 seats, 100 GB and 1.5 GB ingestion, $0.03/GB and $0.10/MB overages, 30-day trial; Enterprise custom with $0.20/MB Weave overage; Academic free with 100 seats (15 October 2026)coreweave.com ↗
  • wandb.ai root redirects to forge.coreweave.com/wandb (a Forge sign-in page) and wandb.ai/site/pricing 301-redirects to coreweave.com/forge-pricing (15 October 2026)forge.coreweave.com ↗
  • W&B marketing site still describes Models, Weave, Training, Inference, Registry, an iOS app and an MCP server, with ISO 27001/27017/27018, SOC 2 and HIPAA (15 October 2026)wandb.ai ↗
  • CoreWeave page presenting Weights & Biases as a CoreWeave company; CoreWeave acquired W&B on 5 May 2025 and announced Forge on 30 September 2026 (15 October 2026)coreweave.com ↗
Fact check

Is Weights & Biases legit?

We don’t label tools safe or unsafe. Below are the records we checked ourselves, each with the date we read it.

Domain registered
Dec 16, 2017 (8 years ago)Registry record (RDAP), registrar 101domain GRS Limited, looked up Sep 6, 2026
TLS certificate
Issued by Google Trust Services, valid until Dec 23, 2026Seen in our own connection to the site, before our last check on Oct 15, 2026
Last checked
Oct 15, 2026The site now redirects elsewhere; we list the tool as acquired
Operated by
CoreWeave, Inc., United StatesOur research, card verified Oct 15, 2026

How we collect these records

Editorial verdict

Pros & cons of Weights & Biases

medium confidence

W&B is still the default experiment tracker for research teams - the SDK is a line of code, the dashboards are the ones everyone has seen in papers, and Weave gives LLM teams tracing and evals in the same account. What has changed is the owner: it is now a feature of CoreWeave Forge, which bundles training and inference on CoreWeave's cloud, and that is the lens through which the roadmap will be set. Teams tied to another cloud should expect the gravitational pull and price the open-source alternatives against the $60-a-month Pro floor.

Pros

  • Integration is one import and a few lines; nearly every framework has a hook
  • Free tier with 5 seats is enough for a small team, and academics get a large free plan
  • Weave adds LLM tracing, evaluation and monitoring in the same workspace
  • Self-hosted and dedicated deployments survive the Forge transition unchanged

Cons

  • Pro bills a $60 minimum plus overages on storage and Weave ingestion, so a chatty tracing setup grows the bill
  • Pro is limited to companies under 50 employees; above that you negotiate
  • The product is now part of a GPU cloud's bundle, and the marketing site, login and pricing are split across wandb.ai and coreweave.com
  • Weave data overage is per MB, which is easy to underestimate for high-volume LLM apps

Best for

  • ML research teams that need run comparison, artifact lineage and shareable reports
  • LLM application teams wanting tracing and evaluation next to their model experiments
  • Organisations already training or serving on CoreWeave

Not for

  • Solo developers who just need local run logs - MLflow or a notebook is free and simpler
  • Teams that must keep their platform cloud-neutral and avoid a GPU vendor's ecosystem

Watch out: The service was Weights & Biases and is now CoreWeave Forge: wandb.ai now redirects to a Forge login, pricing lives on coreweave.com, and the SaaS plans were restructured in September 2026 - check your plan's seat and ingestion caps rather than the ones you signed up with.

#2inMLOps & Experiment Tracking· of 7 tools, ranked by rating and popularity

Be the first to review Weights & Biases

Ask a question about Weights & Biases