- Verified by BizGlocal
AI Developer Platform For Experiment Tracking And LLM Tracing, With Published Support Tiers
- 201-1,000
- 2017
- San Francisco, CA
- 6+
About Weights & Biases
Industry leader in connected fleet operations
Market Position
A widely used developer tool for machine learning teams, now part of CoreWeave. Its homepage says thousands of companies rely on W&B as their system of record. It competes with Arize AI, LangSmith, Comet and MLflow-style tools.
Innovation Focus
Weave for agent tracing, evaluations, scorers and monitors, plus W&B skills and an MCP server so coding agents such as Claude Code can read production data and run evaluations, per W&B.
Core Strengths
One platform for classic ML training and LLM application work, a published three-tier support model, customer-facing training on the Enterprise tiers, and ISO 27001, 27017 and 27018 certifications listed on its homepage.
Use Cases
Experiment tracking, hyperparameter sweeps, model registry and lineage, LLM tracing, prompt and model evaluation, safety and quality scoring, and production monitoring with Slack alerts.
Core Products & Solutions
Weave Tracing
End-to-end observability of agent behavior, with traces organized into sessions, turns, steps, tools and sub-agents. W&B says this works across multi-agent systems.
Weave Evaluations And Scorers
An evaluation API with comparison views and leaderboards. Pre-built scorers cover toxicity, bias, PII detection and hallucinations, plus coherence, fluency and context relevance.
Weave Monitors And Guardrails
Continuous monitoring of production traffic with out-of-the-box signals, and alerts routed to Slack or webhooks. A Playground tests prompts and models against production traces.
W&B Models
Experiment tracking, a centralized model registry and artifacts for classic ML work. W&B describes Registry as a central place for model versions and lineage. The pricing page lists Models under Enterprise.
Sweeps And Reports
Sweeps handle hyperparameter optimization, and Reports are collaborative dashboards that teams share inside or outside the company.
Serverless Inference And Sandboxes
Per-model inference billed monthly, and sandboxes listed on the pricing page. Per-model rates sit on a separate page, so confirm current rates.
Pricing Overview
Free
Individuals and small teams trying W&B or tracking light usage
$0per month
- Up to 5 model seats
- 5 GB storage per month
- 1 GB Weave data ingestion per month
- Support by email and community
- SOC 2 Type II listed for all plans
Pro
Organizations with fewer than 50 employees that need more seats, storage and support
From
$60per month
From $60 per month, billed monthly, with a 30-day free trial
- Up to 10 model seats
- 100 GB storage per month, then $0.03 per GB
- 1.5 GB Weave ingestion per month, then $0.10 per MB
- Email and chat support with 4 to 24 hour first response by severity
- Larger organizations must move to Enterprise
Enterprise
Larger teams that need SSO, security controls, dedicated deployment and higher support tiers
Custom
Pricing
Custom quote, invoiced annually upfront
- Customizable seats and storage, or bring your own bucket
- Weave ingestion at $0.20 per MB, with volume discounts on annual commitments
- Adds SSO, SCIM, audit logs, CMEK, private connectivity and a HIPAA option
- Support first response of 1 to 4 hours by severity and package
- W&B Models is listed on this plan
Who Is Weights & Biases Best For?
Small Businesses
1-25 EmployeesFree and Pro at $60 per month are low entry points, with a 30-day Pro trial. Support is email and community on Free and email plus chat on Pro, so expect no dedicated contact.
Mid-Market Companies
26-100 EmployeesTeams under 50 employees can stay on Pro, but larger ones must move to Enterprise. Standard Plus adds onboarding, a dedicated success team and a Slack or Teams channel, and Enterprise pricing is not published.
Enterprise Companies
100+ EmployeesW&B lists SSO, SCIM, audit logs, CMEK, dedicated deployments and ISO 27001 for this tier. Premium support is required for customer-managed deployments and for customers with two or more deployments.
ML And LLM Teams Under Production Pressure
Any SizeTeams that train models and ship LLM apps benefit from one record of experiments, traces and evaluations. Festo and Graphcore illustrate the experiment-tracking side.
Measurable Impact & Benefits
8.7/10
G2 Quality Of Support
G2 Quality of Support for W&B, shown with 8.6 for Ease of Setup on G2’s comparison page. The response count was not shown.
3
Support Tiers
Standard, Standard Plus and Premium, all described on the W&B support page for Enterprise customers.
1
Hour Premium P0 Response
Premium P0 initial response is 1 business hour, with 2 hours for P1 and 4 hours for P2 and P3, as stated by W&B.
24/7
Premium Coverage
Premium is follow-the-sun support, 24 hours a day, 7 days a week. Standard and Standard Plus run 15 hours a day on weekdays.
8 Hours To 20-30 Minutes
Time Saved
Festo says setting up a new experiment fell from about 8 hours to 20 to 30 minutes with W&B.
50 To 100
Times More Experiments
Graphcore says it runs 50 to 100 times more experiments than before on Mk1 IPU systems.
5-10
To 50-100 Experiments
Gretel says each compute block went from 5 to 10 experiments to 50 to 100, which it describes as 10x experimentation velocity.
10x
Experiment Velocity
Woven Planet says experiment tracking gave it 10x velocity.
Weights & Biases Real reviews
Our Rating:
4.5
- Verified Users
- Verified Users
Compare Alternatives
See how Weights & Biases stacks up against competitors
Get Your Free Checklist
Enter your email to download instantly
No spam. Instant access. Unsubscribe anytime.
Download
Spam
Forever
Weights & Biases’s Pros & Cons
Balanced evaluation to help you make an informed decision
Strengths
- G2 Quality of Support of 8.7 out of 10, with Ease of Setup at 8.6 and overall G2 at 4.5 across 83 reviews
- Three published support tiers with first-response targets from 1 to 24 business hours, as stated by W&B
- Premium adds 24/7 follow-the-sun coverage, quarterly roadmap sessions, early beta access and feature prioritization
- Standard Plus and Premium add Slack or Teams, in-app chat, onboarding and a dedicated success team
- ISO 27001:2022, 27017 and 27018 certifications and SOC 2 listed, with HIPAA and CMEK on Enterprise
- One platform for experiment tracking, a model registry, LLM tracing, evaluations and monitoring
- Named customer results from Festo, Graphcore, Gretel and Woven Planet
Limitations
- Premium and Standard Plus prices are not published, and Premium is required for customer-managed deployments
- Standard support covers email and public forums only, and Free shows email and community
- Standard and Standard Plus hours are 2AM to 5PM Pacific on weekdays, so evenings and weekends are not covered
- Review volume is small: 6 Gartner ratings, 8 PeerSpot reviews and 83 on G2, with no Capterra reviews found
- G2 lists a lower support score than Arize AX, which rates 8.9
- Pro is limited to organizations under 50 employees, and W&B Models is listed on Enterprise
Our Verdict
Weights & Biases earns its Best in Support placement on structure and documentation. It publishes three support tiers with first-response targets by severity, from 4 to 24 business hours on Standard and Standard Plus to 1 to 4 on Premium, plus 24/7 coverage, Slack or Teams channels, onboarding, a dedicated success team and roadmap sessions. G2 Quality of Support is 8.7 out of 10, and enterprise reviewers on PeerSpot describe responsive, technically competent support engineers. The placement has caveats. G2 lists Arize AX at 8.9 for Quality of Support, so W&B does not lead on that score, and review counts are small. The best terms sit in paid tiers whose prices are not published, and Standard hours do not cover nights or weekends. Targets are vendor stated, not audited. Buyers should write the tier, hours and response times into the contract and confirm how support changes under CoreWeave.
Resources
Resources & Documentation
Detailed materials for deeper evaluation
Pricing Page
Free, Pro and Enterprise plans, usage meters, support levels and billing terms
Support Page
Standard, Standard Plus and Premium tiers, hours, response times and channels
Documentation
W&B and CoreWeave documentation, support articles and the docs assistant
Weave Product Page
Tracing, evaluations, scorers, guardrails, monitors and playground
Customer Stories
Case studies from Festo, Gretel, Graphcore, Woven Planet, LG AI Research and others
Festo Case Study
How Festo cut new experiment setup from about 8 hours to 20 to 30 minutes
Gretel Case Study
How Gretel reached 10x experimentation velocity
CoreWeave Acquisition Announcement
CoreWeave’s May 2025 notice of the completed acquisition and its interoperability pledge
What to Know
What to Know Before Buying
-
Standard Plus and Premium add onboarding and a dedicated success team -
Enterprise features such as SSO, SCIM and CMEK need configuration
-
Customer-managed deployments require Premium support
-
W&B lists framework integrations such as Lightning and Trainer on its homepage, and Weave connects to Slack, webhooks and Claude Code through an MCP server. Confirm your exact LLM framework and cloud.
-
Storage, Weave ingestion and inference scale with usage, and Pro overages cost $0.03 per GB of storage and $0.10 per MB of Weave data. Model costs for high-volume tracing before you commit.
-
Standard Plus and Premium support cost extra and vary by deployment and instance size. Weave ingestion, inference and sandboxes are billed in arrears. Storage is averaged over 30 days.
-
Moving from another tracker means re-instrumenting code and recreating dashboards, registries and evaluation datasets. Externally stored data and reference artifacts do not count toward storage quota, per the pricing page.
-
Pro is billed monthly upfront and Enterprise is invoiced annually upfront. Removed Pro seats get no credit. Confirm the support tier, hours, response targets and renewal terms in writing.
Alternatives
Compare Alternatives
See how Weights & Biases stacks up against competitors
Arize’s pricing page lists AX Free at $0 with community support, AX Pro at $50 per month with email support, and AX Enterprise as custom with dedicated support and custom SLAs. Arize also offers the open-source Phoenix tool. Its page names no support tiers or response times, whereas W&B publishes them.
LangSmith’s pricing page lists a Developer plan at $0 per seat, Plus at $39 per seat per month and custom Enterprise with a support SLA, and only Enterprise offers self-hosting. Its pages do not describe support tiers or response times. W&B bundles experiment tracking and a model registry as well.
Get Started
Ready to Get Started?
Get personalized recommendations from top 3 vendors
By submitting, you agree to receive recommendations from verified vendors. No spam, unsubscribe anytime.
Are You a Software Vendor?
List your company and get in front of decision-makers actively searching for solutions.
FAQ’s
Frequently Asked Questions
Common questions about Weights & Biases
Free is $0, Pro starts at $60 per month with a 30-day trial, and Enterprise is custom. Pro includes 100 GB of storage and 1.5 GB of Weave ingestion, with overages billed monthly. Premium support pricing is not published.
W&B lists Standard, Standard Plus and Premium for Enterprise customers. Premium P0 response is 1 business hour with 24/7 coverage, and Standard is 4 business hours. Free has email and community support, and G2 Quality of Support is 8.7 out of 10.
Yes. It is listed in San Francisco, California and was founded in 2017. CoreWeave announced the completed acquisition in May 2025 and says W&B stays interoperable with any infrastructure provider.
Its homepage lists ISO 27001:2022, 27017 and 27018, SOC 2, HIPAA and NIST 800-53 support. Enterprise adds SSO, SCIM, audit logs and CMEK. Request the reports that apply to you.
Weave traces, evaluates and monitors LLM applications and agents. Models covers experiment tracking, sweeps, artifacts and the model registry for classic ML work. Many teams use both.
Yes. CoreWeave announced the completed acquisition on May 5, 2025. Media reported about $1.7 billion, but no terms were disclosed. W&B plans and documentation now appear on CoreWeave’s site.
W&B does not publish a timeline. Starting to log runs or traces is usually quick, and Enterprise security setup takes longer. Standard Plus and Premium add onboarding help.
Common alternatives include Arize AI for LLM observability, LangSmith for LangChain-based apps, Comet for experiment tracking, Langfuse and MLflow, plus Datadog for general monitoring.