What is Weights & Biases?
The AI developer platform for training, tracking, and deploying machine learning models at scale.
- Category
- AI/ML Developer Tooling & MLOps
- Headquarters
- San Francisco, CA
- Founded
- 2017
- Employees
- ~300–311 (post-acquisition)
- Total Funding
- $250M across 5 rounds
- Status
- Acquired by CoreWeave for ~$1.7B (May 2025)
What is Weights & Biases?
Weights & Biases (W&B) is an AI developer platform that provides experiment tracking, model management, and generative AI observability tools used by over 1 million AI engineers at more than 1,400 organizations worldwide — including OpenAI, Meta, NVIDIA, Anthropic, Cohere, and Toyota. Founded in 2017 by Lukas Biewald, Chris Van Pelt, and Shawn Lewis, the company was acquired by CoreWeave (Nasdaq: CRWV) for approximately $1.7 billion in May 2025 after raising $250 million in venture capital and crossing $100 million in ARR. W&B has become the de facto standard for ML experiment tracking across both frontier AI research labs and enterprise AI teams.
W&B's core platform, branded as W&B Models, lets machine learning teams log every training run — hyperparameters, metrics, system resource usage, model checkpoints, and sample predictions — in a centralized dashboard that makes experiments reproducible and comparable across a team. The platform also includes Sweeps for automated hyperparameter search, Artifacts for dataset and model versioning, and a Model Registry for managing the model lifecycle from early experimentation through production promotion. These tools were originally built around classical deep learning workflows but now serve the full spectrum from tabular ML to large language model (LLM) fine-tuning at scale.
In April 2024, W&B launched W&B Weave at its annual Fully Connected conference — a lightweight observability and evaluation toolkit for generative AI applications targeting developers building on foundation models. Weave provides tracing, logging, and online evaluations for LLM pipelines, agents, and RAG systems, extending the platform's reach into the fast-growing LLMOps space. By the time of the CoreWeave acquisition, Weave had become a key growth product as enterprise demand for production AI observability accelerated sharply.
At its peak as an independent company, W&B had logged more than 10 billion API calls daily, integrated with over 100 ML frameworks, and was embedded in more than 20,000 open-source repositories. The company served 70% of the top 10 AI research labs globally and reached over 1 million registered users, cementing its position as the dominant infrastructure layer for AI-native development teams. The CoreWeave acquisition adds a bundling dimension, combining W&B's tooling with CoreWeave's GPU cloud platform to create an end-to-end AI development environment.
What does Weights & Biases offer?
W&B's platform covers the full AI development lifecycle, from classical ML experiment tracking through LLM observability, generative AI evaluation, and managed inference — all accessible via a Python SDK with 50M+ PyPI downloads.
- Experiment Tracking· Core ML
- Hyperparameter Sweeps· Core ML
- Model Registry· Core ML
- Artifact & Dataset Versioning· Core ML
- W&B Weave (LLM Tracing & Eval)· Generative AI
- W&B Prompts· Generative AI
- Online Evaluations· Generative AI
- W&B Inference· Generative AI
- Production Monitoring· MLOps
- Reports & Dashboards· Collaboration
- Self-Hosted / Enterprise Deployment· Platform
- CoreWeave Mission Control Integration· Platform
How does Weights & Biases make money?
W&B operates a seat-based SaaS subscription model with a generous free tier designed to drive individual developer adoption, then expand into paying teams and large enterprise contracts. Revenue comes from three tiers: a free Personal plan, a Pro plan at approximately $50–60/user/month, and a custom-priced Enterprise tier typically ranging from $300–400/seat/month. The company reached $100 million in ARR as an independent company, fueled by a bottom-up product-led growth (PLG) motion that seeds organic adoption across AI teams globally.
The free tier — available for personal and small-team use — provides full access to experiment tracking, Weave tracing, and the model registry with no time limit, seeding adoption at the individual developer level with zero sales friction. The Pro plan (approximately $50–60/seat/month) supports teams and unlocks collaboration features, CI/CD automations, Slack and email alerting, and priority support; additional cloud storage runs approximately $0.03/GB, and Weave data ingestion is metered beyond the included free tier. Pro is explicitly limited to early-stage companies with fewer than 50 employees — teams that exceed this threshold are required to migrate to the Enterprise tier.
Enterprise pricing is negotiated and typically ranges from $300–400/seat/month based on disclosed Vendr contract data — representing a 5–8x uplift over Pro. Enterprise deals include single-tenant cloud or fully self-hosted deployment, HIPAA compliance, SAML SSO, audit logs, customer-managed encryption (BYOK), and a dedicated customer success team. Multi-year enterprise contracts tend to expand as AI teams grow headcount and launch more training workloads, producing strong net revenue retention; W&B has reported top-tier retention rates near 98% for enterprise accounts.
Growth is driven by a classic bottom-up PLG motion: individual ML engineers adopt W&B for free, prove value to their teams, and trigger procurement once collaboration or compliance requirements demand the paid tier. The CoreWeave acquisition introduces a new bundling dimension — W&B tooling is now offered as part of CoreWeave's managed GPU cloud platform through Mission Control and deeper infrastructure integrations, which may shift an increasing share of enterprise revenue toward bundled cloud deals rather than standalone SaaS contracts.
Who leads Weights & Biases?
W&B was co-founded by three engineers — Lukas Biewald, Chris Van Pelt, and Shawn Lewis — who all remain with the company post-acquisition. Biewald serves as General Manager within CoreWeave, while Van Pelt and Lewis hold Distinguished Engineer titles. The VP/C-level layer includes Robin Bordoli (CRO & CMO), Phil Gurbacki (VP of Product), and Mike Saparov (SVP of Technology).
- Lukas BiewaldGeneral Manager & Co-Founder (formerly CEO)2017–presentPreviously co-founded CrowdFlower (acquired by Figure Eight / Appen); Stanford CS graduate who led W&B from founding to a $1.7B exit. Continues to run W&B as a CoreWeave business unit.
- Chris Van PeltDistinguished Engineer & Co-Founder2017–presentPreviously co-founded CrowdFlower with Biewald; deep ML systems background, leads technical architecture and platform strategy.
- Shawn LewisDistinguished Engineer & Co-Founder2017–presentFormer Google software engineer and founder of Beep Networks; de facto CTO during early years, now focused on core platform engineering.
- Robin BordoliCRO & CMO2021–presentHolds dual CRO/CMO role — a signal of the tight PLG coupling between marketing and revenue. Previously CRO/CMO at NextRoll, CEO at Figure Eight, and has IPO experience from Marketo and Jive.
- Phil GurbackiVP of Product2022–presentLeads the product roadmap across W&B Models, Weave, and the broader platform strategy, including the CoreWeave integration roadmap.
- Mike SaparovSVP of Technology2022–presentOversees the core engineering organization and platform reliability at the scale of 10B+ API calls per day.
How do you contact Weights & Biases's leadership?
Weights & Biases uses the first.last@wandb.com email format as the dominant pattern, verified across ContactOut and RocketReach data (approximately 50% of confirmed addresses follow this format). The company's primary published support address is support@wandb.com. Executive emails below follow the verified first.last@wandb.com pattern and are directional outreach addresses — not independently confirmed for each individual.
first.last@wandb.comHow much funding has Weights & Biases raised?
Weights & Biases raised $250 million across five primary venture rounds between 2018 and 2023, reaching a $1.25 billion valuation before being acquired by CoreWeave for approximately $1.7 billion in May 2025 — a strong exit at a 36% premium to the last private round. Key backers include Coatue, Insight Partners, Felicis Ventures, BOND Capital, and individual AI investors Nat Friedman and Daniel Gross.
The company's earliest institutional capital was a $5 million Series A in May 2018 led by Trinity Ventures and Bloomberg Beta, validating the experiment-tracking thesis when the product had fewer than 10,000 users. A Series B followed in two tranches: the first $15 million was led by Coatue in May 2019 as the user base began compounding rapidly through word-of-mouth among ML practitioners; the second $45 million tranche in February 2021 was led by Insight Partners, with GitHub founder Tom Preston-Werner and AI researchers Pieter Abbeel, Nat Friedman, and Richard Socher participating as angels. By the time of the second Series B close, W&B had crossed 70,000 users and 200+ enterprise customers, and cumulative capital raised totaled approximately $65 million.
Velocity accelerated sharply in October 2021 when Felicis Ventures and BOND Capital (associated with investor Mary Meeker) co-led a $135 million Series C at a $1 billion valuation, marking W&B's unicorn milestone. The user base had doubled in 2021 to more than 100,000 active users. In August 2023, Nat Friedman (ex-GitHub CEO) and Daniel Gross (AI investor and former YC partner) led a strategic $50 million top-up at a $1.25 billion post-money valuation, with Sapphire Ventures and all major prior investors participating. Total disclosed capital from these five rounds was $250 million.
On March 4, 2025, CoreWeave (Nasdaq: CRWV) announced a definitive acquisition agreement; the deal closed May 5, 2025 for approximately $1.7 billion. The acquisition price represented a 36% premium over the August 2023 valuation and was driven by strategic logic: CoreWeave needed a native MLOps layer to differentiate its GPU cloud from commodity infrastructure providers. Owning W&B gave CoreWeave an integrated train-deploy-observe platform, justifying a control premium above W&B's standalone SaaS value.
How did Weights & Biases get here?
From a side project to a $1.7B exit in under eight years, W&B tracked the rise of modern deep learning and emerged as the infrastructure layer every serious AI team depends on.
- 2017Company Founded in San FranciscoLukas Biewald, Chris Van Pelt, and Shawn Lewis found Weights & Biases. Early users include OpenAI, Toyota Research, and Uber. The product is born out of frustration with the lack of reproducibility tools in ML research.
- May 2018Series A — $5M (Trinity Ventures, Bloomberg Beta)First institutional capital of $5M closes. Within a year the tracker logs nearly 1 million experiments. Product-market fit confirmed through developer word-of-mouth in ML communities.
- May 2019Series B Tranche 1 — $15M (Coatue)Coatue leads $15M to support rapid developer adoption. Brings total raised to $20M. User base compounding through organic ML community growth.
- Feb 2021Series B Tranche 2 — $45M (Insight Partners)Company crosses 70,000 users and 200+ enterprise customers. GitHub founder Tom Preston-Werner joins as angel alongside AI researchers Pieter Abbeel and Richard Socher. Insight Partners leads.
- Oct 2021Series C — $135M at $1B valuation (Unicorn)Felicis Ventures and BOND Capital co-lead; W&B becomes a unicorn. User base doubles in 2021 to 100,000+. Existing investors Insight Partners and Coatue participate.
- Aug 2023$50M Strategic Round — $1.25B valuation (Nat Friedman & Daniel Gross)Nat Friedman and Daniel Gross lead $50M round. W&B launches W&B Prompts for LLM observability; user base reported at 700,000+. Sapphire Ventures joins the cap table.
- Apr 2024W&B Weave GA Launch at Fully ConnectedW&B Weave, an LLM tracing and evaluation toolkit for generative AI apps, goes generally available at the Fully Connected conference. Extends the platform beyond classical ML into the LLMOps space.
- May 2025Acquired by CoreWeave for ~$1.7BCoreWeave (Nasdaq: CRWV) completes acquisition on May 5, 2025. W&B operates as a business unit inside a publicly-traded AI cloud infrastructure company. Lukas Biewald becomes General Manager within CoreWeave.
Who are Weights & Biases's competitors?
W&B competes across two axes: classical ML experiment tracking (MLflow, Comet ML, ClearML, Neptune) and the newer LLMOps/AI observability space (Arize, LangSmith, Braintrust). Its breadth across both categories is its primary differentiator.
- MLflow (Databricks)Open-source, self-hosted experiment tracking now backed by Databricks — the default for teams in the Databricks/Spark ecosystem, but limited UI polish and no SaaS managed hosting parity with W&B.
- Comet MLClosest commercial alternative to W&B for classical ML teams, with near-identical dashboards and team features; increasingly competing in LLM evaluation as well with Comet Opik.
- ClearMLApache 2.0 open-source with a managed SaaS option; uniquely covers experiment tracking, data versioning, orchestration, and model serving in one self-hostable stack — attractive for cost-sensitive teams.
- Arize AISpecialized in production ML monitoring and LLM observability (via Phoenix open-source); stronger on drift detection and tabular model monitoring, lighter on training-time experiment tracking.
- LangSmith (LangChain)LangChain's native observability and eval platform for LLM apps; narrower scope than W&B Weave but tightly integrated for the very large LangChain developer ecosystem.
- BraintrustEvaluation-first LLM development platform targeting AI product teams; focused on evals and human annotation rather than full model-training lifecycle coverage.
- Neptune.aiClose commercial alternative to W&B for experiment tracking and model registry; strong on metadata management and long-term run history; popular in Europe-based ML teams.
Weights & Biases — frequently asked questions
