Weights & Biases logo
Verified by SaaSOffers
PremiumAI & Data

Weights & Biases Coupon: Free Plan

Free Plan
Verified November 2026

ML experiment tracking, model versioning, and dataset management platform.

✓ Verified deal✓ No spam, ever✓ 10,000+ startups

Deal Highlights

Free Plan
Deal Value
Premium Plan
Access Type
AI & Data
Category

The SaaSOffers Weights & Biases deal

Offer: Free Plan.

Who qualifies: our listing sets no extra requirements.

There is no promo code. You open the deal on SaaSOffers and sign up from there.

We checked these plans and prices against Weights & Biases' official pricing and help pages in November 2026.

Weights & Biases (W&B) tracks machine learning experiments. You add a few lines to your training script, and every run's metrics, hyperparameters, system stats and output files land in a shared dashboard. It also versions datasets and models, runs hyperparameter sweeps, and traces LLM apps through a product called Weave.

Weights & Biases homepage, Free Plan for startups

The Free plan is W&B's standard free tier. Anyone can get it directly on wandb.ai, so this listing does not save you anything beyond what W&B already gives away. W&B is also now part of CoreWeave. CoreWeave completed its acquisition of Weights & Biases on May 5, 2025, and the pricing page at wandb.ai now carries the title "CoreWeave Forge Pricing". The experiment tracking product is listed as "Weights & Biases Models" inside CoreWeave Forge. The W&B name, SDK and docs are still in use.

What W&B does

  • Experiments. Log metrics, configs and system data from each training run, then compare runs side by side.
  • Sweeps. Automated hyperparameter search with Bayesian, grid and random methods, parallelized across one or more machines.
  • Artifacts and Registry. Version datasets and models and track lineage, so you know which data produced which checkpoint.
  • Tables and Reports. Interactive data views and shareable write-ups built from live runs.
  • Weave. Tracing, evaluations, scorers and LLM-as-a-judge metrics for LLM applications.
  • Integrations. W&B's docs list native integrations for PyTorch Lightning, Hugging Face Transformers, Keras and YOLOv5, plus a Python SDK for any other library.

Getting started takes pip install wandb, an API key and a call to wandb.init() in your script.

Weights & Biases plans and features before the Free Plan deal

Free, Pro and Enterprise side by side

PlanPrice billed yearlyBilled monthlyModel seatsStorageWeave ingestionKey limits
Free$0$0Up to 55 GB/month1 GB/monthCommunity support only
ProNot publishedStarts at $60/monthUp to 10100 GB/month, then $0.03/GB1.5 GB/month, extra listed at $0.10/MBCompanies under 50 employees; 30-day free trial
EnterpriseCustom, invoiced annually upfrontCustomCustomizableCustomizable or bring your own bucketExtra listed at $0.20/MB, volume discounts on annual commitmentsSSO, SCIM, audit logs, HIPAA option

W&B's page says Pro "starts at" $60 a month, billed monthly. It does not say on the page whether that is per seat or how the price grows with seats. Check the price for your seat count in the checkout before you commit.

Pro billing rules

  • Pro is billed monthly upfront.
  • Seats added mid-term are prorated. W&B does not issue credits for removed seats.
  • Weave ingestion, serverless inference, sandboxes and storage are billed monthly in arrears.
  • Pro includes $5 a month of CoreWeave Forge Inference credit for a limited time. Extra inference is billed monthly.
  • Pro is meant for teams with fewer than 50 employees. W&B says larger customers must move to Enterprise because Pro has rate limits and performance constraints.

How storage is measured

W&B averages your storage over 30 days. Its own example: 100 GB for 15 days, 200 GB for 1 day and 300 GB for 14 days gives 1,500 + 200 + 4,200 = 5,900 GB-days. Divided by 30, that is 196.6 GB, rounded to 197 GB. Storage includes artifacts and data logged to runs. Files you only reference in your own bucket do not count, only their metadata.

What the Free plan doesn't include

Free covers the full tracking workflow for a small group. It leaves out the team and admin layer:

  • Teams and access control. Pro adds unlimited teams and team-based access controls.
  • Service accounts. Needed to log runs from CI or shared training jobs without a personal key.
  • Automations and alerts. CI/CD automations and Slack and email alerts start on Pro.
  • Support beyond the community. Pro adds priority email and chat support.
  • More storage. Free stops at 5 GB a month. The $0.03 per GB overage is listed for Pro.
  • Weave headroom. Free includes 1 GB of trace ingestion a month.
  • Security features. SSO, SCIM user provisioning, custom roles, audit logs, customer-managed encryption keys, IP allowlisting and the HIPAA option are Enterprise features.

Two worked examples

The Free plan costs $0 whether you sign up through SaaSOffers or directly, so the examples compare Free with the Pro starting price. That gap is what staying on Free saves.

Example 1: a 3-person research team that fits Free

Three ML engineers track experiments for one model. They log about 4 GB of runs and checkpoints a month and do not use Weave.

  • Model seats: 3, within Free's limit of 5
  • Storage: 4 GB, within Free's 5 GB
  • Cost on Free: $0
  • Cost on Pro at the starting price: $60 x 12 = $720 a year
  • Saving from staying on Free: $720 a year

The limit to watch is storage. Once checkpoints push the 30-day average past 5 GB, the team needs to delete old artifacts, keep large files in its own bucket as references, or upgrade.

Example 2: a 6-person startup that needs Pro

Six people means more than Free's 5 model seats, so this team needs Pro. They store about 250 GB of artifacts and send 3 GB of Weave traces a month from an LLM feature.

  • Pro base: from $60 a month (seat pricing not published, so treat this as a floor)
  • Storage overage: (250 GB - 100 GB) x $0.03 = $4.50 a month
  • Weave overage: (3 GB - 1.5 GB) = 1,500 MB x $0.10 = $150 a month
  • Monthly total at least: $60 + $4.50 + $150 = $214.50
  • Yearly total at least: $214.50 x 12 = $2,574

Weave ingestion is the expensive meter here, not storage. At $0.10 per MB, every extra gigabyte of traces costs about $100. Sample traces in production and log only what you will review. The $5 monthly inference credit is too small to change this math.

The rest of CoreWeave Forge

Since the acquisition, W&B's site sells experiment tracking next to a wider set of CoreWeave products. None of these are part of the Free plan offer, but you will see them in the menu and on the bill if you turn them on:

  • Serverless SFT and Serverless RL. Fine-tuning and reinforcement learning on open models without provisioning GPUs.
  • Model Distillation. In preview. It trains a smaller model and scores it against the one you run.
  • Serverless Inference. Open models on one API, billed per token. Dedicated Inference puts your own weights on GPUs you choose.
  • Agent Lens. In preview. Observability for agent traces, alongside Weave.
  • Sandboxes. Isolated environments for agent and RL code, billed on the CPU and memory each sandbox requests.
  • ARIA. A research agent that analyzes your runs, offered for a limited time on the pricing page.
  • Notebooks. Reactive Python notebooks connected to your runs.

The pricing page bills inference, sandboxes, Weave ingestion and storage monthly in arrears. Seats are the only part paid upfront on Pro.

What Weave covers

Weave is the LLM side of the product. The pricing page lists tracing of inputs, outputs and metadata for each inference, evaluations that compare fine-tuning, RAG, LLM and dataset recipes on accuracy, latency and token usage, LLM-as-a-judge metrics, production monitoring of cost and health, and PII redaction. Traces count toward the monthly ingestion allowance: 1 GB on Free and 1.5 GB on Pro. W&B suggests using the free trial to estimate how much ingestion your app produces, since it varies by use case.

W&B defines ingested bytes as data it receives, processes and stores for you, such as trace metadata and LLM inputs and outputs. HTTP headers and other overhead are not counted, and each byte counts once, when it is stored.

Academic license

Students, professors and postdocs doing research unrelated to a for-profit company can get a free Pro license. It includes Pro features, unlimited tracked hours, 200 GB of cloud storage, up to 25 GB a month of Weave ingestion and up to 100 seats. Extra storage costs $0.03 per GB. An active academic email is required. If you qualify, it is worth far more than the Free plan.

Deployment and security

W&B offers three ways to run it:

  • Multi-tenant cloud (now part of CoreWeave Forge), hosted on Google Cloud. This is the self-serve option W&B recommends for trials and for teams that don't need a private deployment.
  • Dedicated Cloud, isolated infrastructure run by W&B on AWS, Google Cloud or Azure, for data residency and stricter governance.
  • Self-Managed, where you run W&B Server on AWS, Google Cloud, Azure or on-premises bare metal and handle upgrades and maintenance yourself.

W&B lists SOC 2 Type II. HIPAA compliance is possible with a bring-your-own-bucket setup. IP allowlisting requires private hosting. Customer-managed encryption keys are available on AWS and Google Cloud.

Who it's for

  • Small ML teams of up to 5 people training models in PyTorch, Keras or Hugging Face Transformers who want shared run history without running a server.
  • Researchers who need sweeps and run comparison. Academics should take the free Pro license instead.
  • Startups with an LLM feature that want tracing and evals next to training runs, as long as trace volume stays near 1 GB a month.
  • Teams that may grow into Enterprise and want SSO, audit logs and private hosting later without changing tools.

Who should skip it

  • Teams that want to self-host for free. W&B's Self-Managed option is arranged with W&B, and some features need an Enterprise license. MLflow is open source under Apache 2.0.
  • Heavy LLM tracing workloads on a budget. Weave overage at $0.10 per MB on Pro adds up fast.
  • Companies with 50 or more employees. Pro is not meant for you, and Enterprise pricing is a sales call.
  • Anyone who wants a discount on Pro. This listing is the free tier, not a paid-plan deal.

W&B vs alternatives

ToolFree optionPaid entryHostingNotes
Weights & BiasesFree: 5 model seats, 5 GB storage, 1 GB WeavePro from $60/monthCloud; Dedicated or Self-Managed on EnterpriseNow part of CoreWeave Forge
MLflowFree and open source under Apache 2.0None for the open-source projectYou host itDescribes itself as an open source AI engineering platform for agents, LLMs and ML models
Comet (Opik)Free Cloud: up to 10 team members, 25k spans a month, 60-day retentionPro Cloud $19/month: up to 50 members, 100k spansCloud or open sourceExperiment tracking is the separate MLOps product: Free for 1 user, Pro $19 per user/month for up to 10 users; extra Opik spans $5 per 100k

Comet's Opik Pro at $19 a month is cheaper than W&B Pro's $60 starting price for LLM tracing. For teams, compare per user: Comet MLOps Pro is $19 per user, while W&B does not say whether its $60 is per seat. For classic experiment tracking, Comet's MLOps Pro plan costs $19 per user a month for up to 10 users with 1,500 training hours, and its MLOps Free plan covers 1 user. MLflow costs nothing in licence fees, but you pay for the server, database and storage you run it on.

Setup tips

  1. Create a team early. Personal projects are fine for one person, but moving runs later takes effort.
  2. Log references, not files. Point artifacts at your own S3 or GCS bucket so large checkpoints don't count against the 5 GB.
  3. Set a run naming convention. Use config in wandb.init() for every hyperparameter so comparisons and sweeps work.
  4. Start sweeps small. Random or Bayesian search over a few parameters before a large grid.
  5. Sample Weave traces. Ingestion is the meter most likely to push a team off Free.
  6. Use the 30-day Pro trial to measure your storage and Weave use before you pay.

How to claim the deal

  1. Open the Weights & Biases deal on SaaSOffers and follow the link to W&B.
  2. Sign up for a free account. No code is needed.
  3. Install the SDK with pip install wandb and create an API key in your settings.
  4. Call wandb.init() in your training script and log your first run.
  5. Invite up to 4 teammates to stay within the 5 model seats on Free.

Who Is This Deal For?

Early-Stage Startups

Seed and pre-seed companies looking to move fast without overspending on tools.

Growing SaaS Teams

Series A+ companies scaling their stack and optimizing software costs.

Solo Founders

Indie hackers and bootstrapped founders who need enterprise tools at startup prices.

Get Free Plan off Weights & Biases

Premium deal. Upgrade once, unlock everything.

Sign Up & Claim

Frequently Asked Questions

Everything you need to know about this startup deal.

Get the weekly deals digest

New verified startup deals every week. No spam, ever. Unsubscribe anytime.

Related Offers