Snorkel AI pricing

About

Snorkel AI is an AI data-development company that began as a research project in the Stanford AI Lab in 2015 and spun out as a company in 2019. Founded by Alexander Ratner (CEO), Chris Ré, Henry Ehrenberg, Braden Hancock, and Paroma Varma, it grew out of more than half a decade of academic work on weak supervision and programmatic labeling — the idea that you can encode domain expertise as labeling functions and auto-generate training data instead of hand-annotating it, backed by 60+ peer-reviewed publications.

The commercial product line has three pillars. Snorkel Flow is the flagship AI data-development platform, where teams build and refine training and evaluation datasets programmatically and fine-tune models on proprietary data — Snorkel claims this is “10-100x faster” than manual annotation. Snorkel Evaluate (GA in May 2025) handles model and agent evaluation. Snorkel Expert Data-as-a-Service supplies expert-curated and frontier-grade datasets directly to AI labs and enterprises. Customers cited include Apple, Google, Intel, Uber, LinkedIn, BNY Mellon, Stanford Medicine, and “5 out of the top 10 US banks.”

Snorkel raised a $100M Series D at a $1.3B valuation in May 2025 (led by Addition, with investors including BlackRock and GV), bringing total funding to roughly $238M. Third-party trackers report 2025 revenue of about $148M ARR, up sharply from $36.8M in 2024 — a jump driven largely by the new expert-data business. For current details, see Snorkel’s site.

Pricing summary : How Snorkel AI’s pricing model works

Snorkel AI is enterprise sales-only. There is no public self-serve rate card for any commercial product — the website routes every pricing intent to “Request dataset samples,” “Talk to a data researcher,” or “Talk to a strategist.” Pricing is bespoke, structured as annual platform subscriptions and committed contracts, and splits across three surfaces:

1. Snorkel Flow / Snorkel Evaluate — annual platform subscriptions for the data-development and evaluation platform, priced against user count and data volume, with cloud or on-premise deployment options.
2. Snorkel Expert Data-as-a-Service — committed data-licensing contracts for expert-curated datasets, scoped by volume and domain rather than a per-record meter.
3. Open-source Snorkel library — free on GitHub, but a research toolkit, not the managed platform.

Snorkel publishes no prices, so the only concrete figures are third-party and indicative: an AWS Marketplace hosted-application listing showed a 12-month contract at $60,000, and one industry guide estimated entry contracts starting around $50,000 per year. Real enterprise deals are widely described as six-figure annually. Treat both as outside estimates, not Snorkel’s official numbers.

What makes this different: Snorkel sits on an open-core split — the weak-supervision library that built its reputation is free, while the commercial platform and expert-data contracts are fully gated. And unlike usage-metered AI infra, the contract value scales against usage-like dimensions (users, data volume, expert-data units) but is negotiated up front as a committed annual deal, not billed per token or per record.

Pricing by product

Snorkel publishes no rate card; the table below reflects the model and only third-party-attributed indicative figures (estimates, not official prices):

Indicative third-party figures (NOT an official rate card): an AWS Marketplace hosted-application 12-month contract listed at $60,000; one industry guide estimated entry deals starting around $50,000 per year; enterprise contracts widely described as six-figure annually.

Sales motions across products: every commercial product is sales-led — discovery call, scoping, and a custom quote. There is no self-serve checkout and no PLG funnel; the open-source library is the only no-cost on-ramp.

Hidden costs : What Snorkel AI users actually pay

Because Snorkel is contract-priced, the “hidden” costs are less about overage fees and more about scope creep and the people you need around the platform:

The biggest real-world cost driver isn’t a line on Snorkel’s invoice — it’s total cost of ownership. As one pricing guide notes, the license is only part of the spend; getting value out of programmatic labeling typically requires dedicated data scientists and ML engineers, which can add multi-hundred-thousand-dollar annual costs before results land. The second driver is scope: because everything is negotiated, user count, data volume, deployment model, and the amount of expert data all move the number, so the same logo can pay very different amounts year over year.

Want to estimate your own Snorkel AI bill? Use the Snorkel AI pricing calculator to model your costs based on users, data volume, and deployment.

Pricing evolution : Snorkel AI pricing history and changes

Cadence

Tracked range: 2019–present. No public price points exist to track; evolution is by product line and contract model rather than a rate card.

Notable changes

• 2021 — Snorkel Flow established as the enterprise platform, sold via annual subscription / committed contract following the $85M Series C at a $1B valuation. No public pricing from the outset.
• 2024 — Third-party references point to an AWS Marketplace hosted-application 12-month contract at $60,000 and entry deals around $50,000 per year, with six-figure enterprise contracts the norm.
• May 2025 — Snorkel raised a $100M Series D at a $1.3B valuation and took Snorkel Evaluate and Snorkel Expert Data-as-a-Service to GA, adding committed expert-data / data-licensing contracts on top of the platform subscription. Pricing stayed sales-only.

The direction of travel is a shift from selling only a labeling platform toward also selling the data itself — Expert Data-as-a-Service moves Snorkel into committed data-licensing deals with frontier labs, which is where most of the 2024-to-2025 ARR growth came from.

What’s unique : Snorkel AI’s distinctive pricing mechanics

1. Open-core, fully-gated commercial. The weak-supervision library that built Snorkel’s reputation is free OSS, but the entire commercial line — Flow, Evaluate, Expert Data — is sales-only with no published price. The free tier is a research toolkit, not a funnel into self-serve.

2. Selling data, not just software. Expert Data-as-a-Service prices committed data-licensing contracts for expert-curated datasets — closer to a frontier-lab data supplier than a SaaS seat. Value is scoped by data volume and domain expertise, not a per-record meter.

3. Contract value scales on usage-like dimensions without metering. Deals flex against users, data volume, and deployment, but are negotiated up front as committed annual subscriptions — committed economics with usage-shaped sizing, the opposite of pay-as-you-go.

Strengths & weaknesses

Billing UX : Snorkel AI billing controls and transparency

• Billing controls — Committed annual subscriptions and data-licensing contracts negotiated with sales; no in-product purchase, metering dashboard, or pay-as-you-go option for the commercial platform.
• Usage visibility — The platform exposes data-development and evaluation workflows, but there is no public spend meter or self-serve usage-to-bill view — consumption is governed by the contract envelope, not real-time metering.
• Payment options — Enterprise invoicing and committed contracts; available via direct sales and (for some hosted units) AWS Marketplace private offers, billed annually.

Strategic wins : Why Snorkel AI’s pricing decisions worked

1. Open-core trust, gated value capture

By keeping the weak-supervision library free and open while gating the platform, Snorkel earned developer credibility (60+ papers, OSS adoption) and still captured enterprise value behind a sales-led wall. See how AI companies structure pricing.

2. Moving up the stack into expert data

Launching Expert Data-as-a-Service let Snorkel sell committed data-licensing contracts to frontier labs — a higher-value, scarcer good than labeling software — which helped drive ARR from $36.8M to roughly $148M in a year. Related: outcome-based pricing trends.

3. Value-priced contracts backed by ROI proof

Anchoring six-figure contracts to “seven-to-eight figure ROI” stories (a biotech saving $10M on data extraction) justified committed pricing to enterprise buyers without ever publishing a rate. See choosing the right usage metric.

Areas to improve : Gaps in Snorkel AI’s pricing approach

1. No transparency for evaluating buyers

A complete absence of public pricing forces every prospect into a sales call, which slows the early buyer journey and pushes price-sensitive teams toward self-serve alternatives. Even a “starting at” band would reduce friction. See bill shock and cost unpredictability.

2. No mid-market on-ramp

The jump from free OSS to six-figure enterprise contracts leaves no paid path for smaller teams. A scoped, self-serve tier of Snorkel Flow could capture demand that currently never reaches sales.

3. Forecastability of committed data contracts

Because Expert Data-as-a-Service is scoped per engagement, customers can struggle to forecast multi-year spend. Clearer volume tiers or published data-unit bands would make budgeting easier for repeat buyers.

Key takeaways

1. Snorkel AI is enterprise sales-only — Snorkel Flow, Evaluate, and Expert Data-as-a-Service are annual committed contracts with no public rate card. For the underlying model, see the introduction to usage-based pricing.
2. It’s open-core: the OSS weak-supervision library is free, but the commercial platform is fully gated with no free tier and no self-serve trial.
3. Indicative third-party figures only — roughly $50,000/yr entry and a $60,000 AWS Marketplace 12-month contract — with real enterprise deals in six figures; these are outside estimates, not Snorkel’s numbers.
4. The business is shifting from software to data — Expert Data-as-a-Service (data-licensing for frontier labs) drove most of the 2024-to-2025 ARR jump from $36.8M to ~$148M.
5. Pricing scales on usage-like dimensions but stays committed — users, data volume, and deployment move the number, yet deals are negotiated up front, not metered.

UBP implications

1. Usage-shaped sizing without usage-based billing. Snorkel shows you can let users, data volume, and deployment drive contract value while still selling committed annual subscriptions — useful when buyers want budget certainty over pay-as-you-go.
2. Data itself can be the metered good. Expert Data-as-a-Service reframes “usage” as data-licensing units, a model any vendor sitting on proprietary or expert data can borrow.
3. Open-core lowers evaluation friction even when the paid product is gated. A free library does the trust-building that a missing public price can’t — a reusable pattern for sales-led AI infrastructure.

Sources

• Snorkel AI official website — no public pricing, contact-sales only (accessed 2026-06-15)
• Snorkel Flow product page (accessed 2026-06-15)
• eesel AI — A complete guide to Snorkel AI pricing in 2025 — AWS Marketplace $60,000 contract, ~$50,000/yr estimate (accessed 2026-06-15)
• Contrary Research — Snorkel AI Business Breakdown — founding, customers, ROI claims (accessed 2026-06-15)
• Snorkel AI Announces $100M Series D and Expanded Platform — Evaluate + Expert Data-as-a-Service GA (accessed 2026-06-15)
• GetLatka — Snorkel AI revenue & valuation — $148M ARR 2025, $36.8M 2024 (accessed 2026-06-15)

Bottom line

Snorkel AI is a textbook open-core, sales-only enterprise vendor: the weak-supervision library that made its name is free, while Snorkel Flow, Snorkel Evaluate, and Expert Data-as-a-Service are sold as committed annual contracts with no published price. The only concrete numbers are third-party and indicative — roughly $50,000/yr entry and a $60,000 AWS Marketplace contract — with real enterprise deals running into six figures and scaling on users, data volume, and the amount of expert data. The strategically interesting move is the shift from selling labeling software to selling the data itself, which powered an ARR jump from $36.8M to roughly $148M into 2025. Browse the pricing blueprint for more fully-researched company profiles, or compare Snorkel against other data-platform companies.

Want to compare Snorkel AI against other AI data and infrastructure companies? Browse the pricing blueprint.