Nomic pricing

About

Nomic, Inc. is an AI company best known for its open-weight Nomic Embed text and multimodal embedding models and the Atlas data-exploration platform for visualizing, searching, and curating large unstructured datasets. As of mid-2026 the company’s primary commercial motion has shifted toward a domain-specific AEC (architecture, engineering, and construction) product: the Nomic Platform deploys agentic workflows over a firm’s drawings, specs, and project data to automate code-compliance checks, drawing review, submittal review, and firm-wide search.

The company sells two distinct things under one brand. The Nomic Platform (priced at www.nomic.ai/pricing) is a seat-based enterprise product aimed at design and construction firms, while the Atlas data-exploration app (priced separately at atlas.nomic.ai/pricing) remains a self-serve tiered subscription for data scientists and ML teams. The same Parse, Extract, and Embed models that power the Platform are also exposed through a Developer API. Nomic, founded in 2022 and backed by Coatue and others, built its early reputation on open-weight research — the open-source Nomic Embed models remain Apache 2.0 — before steering its commercial motion toward the vertical AEC market in late 2025.

That straddle matters for pricing strategy: Nomic now competes in two different markets at once. As a vertical platform it sits alongside enterprise construction-tech tooling, where buyers expect seat-based pricing with annual commitments; as an embeddings provider it competes with Cohere, Voyage AI, and OpenAI on per-token cost, where the relevant comparison is the metered token-based meter. The result is a pricing surface that genuinely reads as two companies bolted together.

Pricing summary : seat-plus-platform-commit with a pooled AI allowance and per-token embedding overage

Nomic runs a hybrid model that differs by surface. The flagship Nomic Platform combines a seat charge with a platform commitment and a pooled, metered AI-usage allowance: $40/user/month on annual commitment, a 25-seat minimum, and a $1,000/month minimum platform commitment. Each seat contributes $20 of “included AI usage” to an org-level pool that is consumed across Platform, Assistant, Workflows, and Developer API; overages only begin after that pool is exhausted. The separate Atlas app is a more conventional tiered subscription with explicit per-token embedding overage.

The billing dimensions captured:

• Seats — $40/user/month (Nomic Platform), $125/seat/month (Atlas Business), annual on the Platform.
• Platform commitment — $1,000/month minimum (Nomic Platform), bundles the first 25 seats.
• Included AI usage pool — $20 per seat per month (Nomic Platform), pooled org-wide, rolls forward within the annual term.
• Per-token embedding usage — on Atlas, $1 per 10M text tokens and $1 per 50k images charged as overage after each tier’s included monthly pool.
• Assistant queries — metered per tier (3 / 30 / 1,000-per-seat) on Atlas.

What makes this different: the Nomic Platform decouples access (seats) from consumption (a single org-wide AI-usage pool funded $20 at a time per seat), so adding a seat raises both the headcount cap and the shared usage budget simultaneously.

Pricing by product

Nomic Platform (AEC) — Business & Enterprise

Platform-commitment mechanics (from the pricing page’s “How pricing works”): the $1,000/month minimum bundles 25 seats on an annual commitment; each $40 seat adds $20 to a pooled AI-usage balance shared across Platform, Assistant, Workflows, and Developer API; overages occur only after the included pool is consumed, with alerts and limits in the admin dashboard. Additional seats are $40/user/month. Bulk data indexing is a guided, separately scoped process by default (Enterprise unlocks self-guided bulk indexing).

Atlas (data-exploration app) — self-serve tiers

Nomic Embed — per-token embedding rate

The Embed Text endpoint (POST https://api-atlas.nomic.ai/v1/embedding/text) serves nomic-embed-text-v1 and nomic-embed-text-v1.5 (8,192-token max input; task types search_document, search_query, classification, clustering). The docs API reference does not publish a standalone dollar-per-token rate; the captured per-token embedding price is the Atlas overage rate of $1 per 10M text tokens (and $1 per 50k images), charged after each Atlas tier’s included monthly pool. On the Nomic Platform, embedding consumption instead draws from the org-level $20-per-seat included AI-usage pool. The new AEC-focused Developer API overview page (www.nomic.ai/developer) lists “Contact Sales” with no public per-call price.

Sales motions across products: PLG / self-serve for Atlas Starter/Plus/Business and the Nomic Embed API; sales-led for the Nomic Platform (Business and Enterprise) and Atlas Enterprise.

Hidden costs : the 25-seat floor, the platform commitment, and embedding overage past the pool

The biggest “hidden” cost on the Nomic Platform is structural: you cannot buy fewer than 25 seats, and the $1,000/month platform commitment applies regardless. A small pilot is not an option at list price.

Archetype A — a 30-person AEC firm onboarding the Nomic Platform. They license 30 seats but exceed the included AI-usage pool in heavy code-compliance months.

The 30 seats clear the $1,000 platform floor, so the marginal cost is the $400 of AI-usage overage once the org burns through its $600/month pooled allowance. The lesson: at this size the seat fee dominates and overage is a manageable tail — but a 10-person firm would still pay the 25-seat minimum ($1,000/month) for capacity it cannot use.

Archetype B — a data-science team using Atlas Business for private embeddings. Five seats, heavy text embedding.

The per-token rate is cheap, so on Atlas the seat fee dominates and embedding overage is a rounding error unless volumes are extreme — the opposite of a pure consumption model where the meter drives the bill. Teams running very high embedding volumes are often better off self-hosting the open-weight Nomic Embed models for free.

Want to estimate your own Nomic bill? Use the Nomic pricing calculator to model your monthly cost based on seats, the included AI-usage pool, and embedding token/image overage.

Pricing evolution : from open embeddings (2024) to a vertical AEC platform (late 2025)

Nomic’s pricing history is really a positioning history. For its first two years the company was an open-research embeddings and data-visualization shop; its monetizable surface was the Atlas app and a metered embedding API. In late 2025 the primary site repositioned to a seat-priced AEC platform, and the embeddings story moved to a secondary domain.

Cadence

Tracked range: 2024-02–2026-06. Quarters not listed were verified stable (0 price changes, 0 SKU additions) in available archives. Atlas tier dollar figures predate the late-2025 pivot but archived snapshots render as a JS SPA, so exact historical figures are not independently legible; current figures are from the 2026-06-03 live capture.

Notable changes

• 2024-02-01 — Nomic Embed text-v1 launches as the first fully reproducible open embedding model to beat OpenAI’s ada-002 on MTEB (nomic.ai/news); it anchors the metered $1/10M-token Atlas embedding rate still in effect.
• 2024-06 — Nomic Embed Vision extends the meter to images ($1 per 50k) by sharing the text latent space.
• 2025-02-17 — Nomic Embed Text v2 ships as an open multilingual MoE model, keeping the open-weight posture even as the company moves upmarket.
• 2025-11 — The AEC pivot: the Nomic Platform replaces the embeddings-first homepage with a vertical construction product priced on seats plus a $1,000/month platform commitment and a pooled AI-usage allowance.

The embeddings-to-AEC pivot in detail

The most consequential change is not a price move but a market move. Before late 2025, Nomic’s commercial pricing was essentially Atlas tiers plus a per-token embedding meter — a developer/data-science motion. After the pivot, the headline product is a vertical AEC platform sold to architecture and construction firms with a 25-seat minimum, an annual commitment, and a $1,000/month floor. The embedding meter did not disappear; it was re-packaged. On the Platform, embedding and indexing consumption now draws from the org-level $20-per-seat included AI-usage pool rather than being billed per token. The same models, the same underlying meter — but wrapped in a seat-priced enterprise commitment and aimed at a non-developer buyer. This is a textbook case of a company moving from a bottom-up usage meter to a top-down seat-and-commit packaging to chase enterprise contract values.

What’s unique : pooled AI usage funded per seat, an open-weight model under a paid platform, and a dual-surface split

Half of every seat fee is prepaid usage. Each $40 Nomic Platform seat includes $20 of AI usage that flows into a single org-wide pool. This is unusual: most seat-priced products treat usage as a separate overage line, but Nomic explicitly funds the meter from the seat fee, so adding a seat raises both the headcount cap and the shared consumption budget at the same time. It blurs the line between seat-based and usage-based billing deliberately.

A pooled, org-level meter rather than per-seat metering. AI usage is not metered per user — it aggregates into one balance consumed across Platform, Assistant, Workflows, and the Developer API. Heavy users and light users share a common pool, which simplifies admin forecasting and avoids per-seat overage surprises.

Metadata-only visibility means no charge until you act. Connected SharePoint, Egnyte, ACC, or Bentley sources surface metadata for free; billing only begins when you explicitly index or invoke AI. This decouples connecting data from paying for it — a meaningful trust signal for firms with terabytes of legacy project archives.

An open-weight model living under a paid platform. Nomic Embed remains Apache 2.0 and self-hostable for free, yet the same models are the engine of a paid seat-and-commit platform. Few companies run an open-source flagship and a closed enterprise commitment in parallel; Nomic’s bet is that the open model drives credibility and the platform captures the enterprise spend.

One brand, two pricing philosophies. The Atlas app is a conventional self-serve freemium ladder; the Platform is a sales-led seat-plus-commit enterprise contract. The same company runs both motions simultaneously against different buyers.

Strengths & weaknesses

Billing UX : pooled-usage admin controls, alerts, and metadata-only data visibility

Named billing controls captured from the pricing page and FAQ:

• Org-level AI-usage pool — the $20-per-seat allowance aggregates into a single shared balance across Platform, Assistant, Workflows, and Developer API; adding a seat raises the pool by $20.
• Admin dashboard alerts & limits — overages “only occur after included AI usage is consumed,” with alerts and spend limits set in the admin dashboard.
• Rollover within the annual term — unused included AI usage rolls forward within the annual commitment period.
• Metadata-only visibility (no charge until indexed) — users see metadata from connected SharePoint, Egnyte, ACC, or Bentley sources but are billed only when they explicitly index or invoke AI.
• Centralized team billing & usage analytics — included on the Business plan; usage analytics and reporting surface per-org consumption.
• Add-on usage packages — teams can buy additional included-AI-usage packages, or add seats ($40/user/mo) to raise the pool.
• Side-by-side plan comparison — the pricing page renders a Business vs Enterprise feature matrix (custom AI commits, SCIM, audit logs, custom deployment, etc.).

Strategic wins : funding the meter from the seat, pooling usage, and using an open model as a wedge

1. Funding the usage meter from the seat fee removes overage anxiety

By including $20 of AI usage inside every $40 seat, Nomic gives buyers a built-in consumption budget before any overage applies. This is a smart answer to the classic enterprise objection to usage-based pricing: unpredictability. The seat fee feels like a flat subscription while the underlying meter still aligns cost to value, much like the credit-pool hybrids other AI tools have adopted.

2. Pooling usage at the org level beats per-seat metering for forecasting

Nomic aggregates the included allowance into one shared balance rather than per-user quotas. Heavy and light users net out against a single pool, so finance teams forecast one number instead of policing dozens of seats. This sidesteps the “dead seats” problem that plagues strict per-seat metering and makes the platform commitment easier to defend in budget reviews.

3. The open-weight model is a credibility wedge for an enterprise platform

Nomic Embed stays Apache 2.0 and self-hostable, which earns developer trust and benchmark credibility — then the same models power a paid seat-and-commit platform. The free model is the top of a funnel whose bottom is a $12,000/year-minimum enterprise contract. Few firms execute this open-core-to-vertical-platform move; Nomic’s is unusually clean because the open artifact and the paid product share an engine.

4. Metadata-only visibility lowers the activation barrier for data-rich firms

Letting customers connect terabytes of SharePoint/Egnyte/ACC/Bentley data and see metadata for free — charging only on explicit indexing — removes the scariest line item from the first conversation. It converts “how much will all our data cost?” into “pay only for what you actually use,” a land-and-expand motion well suited to large legacy archives.

Areas to improve : the small-firm floor, the dual-surface confusion, and the missing API price

1. The 25-seat / $1,000-month floor excludes small AEC firms

The list-price entry point is effectively $12,000/year for capacity a 5-to-10-person firm cannot use. A genuine starter tier — say a 5-seat “Studio” plan with a proportionally smaller commitment — would open the long tail of boutique architecture practices that are exactly the buyers most starved for automation. The fix is a smaller committed package, not a free trial.

2. The two-surface split confuses the core pricing question

“What does Nomic cost?” has no single answer because the Platform and Atlas live on different domains with different models. A unified pricing hub that explicitly frames “Platform for AEC firms / Atlas for data teams / Embed API for developers” — instead of two disconnected pricing pages — would cut buyer confusion and reduce wasted sales cycles. Clear pricing-page information architecture is itself a conversion lever.

3. No public Developer API price undercuts the open-model goodwill

The www.nomic.ai/developer page routes to “Contact Sales” with no per-call rate, which sits awkwardly next to an open-weight, transparently-benchmarked model. Publishing a per-token Developer API rate (as it already does for Atlas overage) would let developers self-serve and would convert the open-model audience into paying API customers without a sales conversation.

4. Platform embedding cost is opaque relative to Atlas

On Atlas the embedding meter is published ($1/10M tokens); on the Platform the same consumption silently draws from the pooled allowance with no visible per-unit rate. Surfacing the effective per-document or per-token conversion inside the admin dashboard would let Platform buyers reason about marginal cost the way Atlas users already can.

Key takeaways

1. Bundle the meter into the seat to neutralize overage fear. Nomic funds $20 of usage from each $40 seat, so the bill reads like a subscription while the underlying model still meters consumption — a pattern any team adding usage to a seat product can copy.
2. Pool usage at the org level, not per seat. A single shared balance is far easier to forecast and defend than dozens of per-user quotas, and it eliminates the dead-seat waste problem.
3. An open-weight flagship can be a paid platform’s funnel. Nomic Embed’s Apache 2.0 license drives credibility and developer adoption that feed a seat-and-commit enterprise product — open-core works as a wedge, not just charity.
4. Charge on action, not on connection. Metadata-only visibility (pay only when you index) removes the scariest cost from the first sales conversation for data-rich buyers.
5. A pivot is a packaging decision, not just a product one. Nomic kept the same embedding engine but re-packaged it from a per-token meter into a seat-plus-commit enterprise contract to chase larger AEC deal sizes.

UBP implications

1. Seat-funded usage pools are a viable bridge from subscription to consumption. By pre-funding the meter from the seat fee, Nomic gets the alignment of usage pricing with the predictability buyers demand — a template for SaaS teams nervous about pure metering.
2. Org-level pooling changes the unit of forecasting. When usage is pooled rather than per-seat, the value metric shifts from “users” to “organizational consumption,” which is a more honest proxy for value delivered in agentic, automation-heavy products.
3. Open models and metered APIs can coexist with sales-led platforms. Nomic shows that a single company can run a freemium open meter and a sales-led commitment side by side — usage pricing need not be all-or-nothing across a product portfolio.

Sources

• Nomic Platform pricing — www.nomic.ai/pricing (accessed 2026-06-03)
• Atlas app pricing — atlas.nomic.ai/pricing (accessed 2026-06-03)
• Nomic Developer API overview — www.nomic.ai/developer (accessed 2026-06-03)
• Nomic Embed Text API reference — docs.nomic.ai (accessed 2026-06-03)
• Nomic News (announcements / changelog) — www.nomic.ai/news (accessed 2026-06-03)
• Nomic Embed Text v1 announcement — www.nomic.ai/news/nomic-embed-text-v1 (accessed 2026-06-03)

For the full corpus of analyzed pricing models, see the pricing blueprint index.

Bottom line

Nomic is two pricing stories in one brand: an open-weight embeddings lineage that gave it credibility, and a late-2025 pivot to a seat-priced AEC platform — $40/user/month with a 25-seat minimum, a $1,000/month commitment, and a clever $20-per-seat pooled AI-usage allowance — that is where the company now chases enterprise revenue. The mechanics are genuinely interesting (funding the meter from the seat, pooling usage org-wide, charging only on indexing), but the small-firm floor and the two-surface confusion are real friction. See how it compares across the pricing blueprint.

Pricing timeline : Major events on a vertical axis

Each milestone below corresponds to a public pricing change, product launch, or material adjustment. Major events use a filled marker; minor adjustments use a faded one.

Two surfaces confirmed — AEC Platform + Atlas tiers

Jun 2026

As of the 2026-06-03 capture, www.nomic.ai/pricing lists Business $40/user/month (25-seat min, $1,000/month commit, $20 included AI usage/seat) and custom Enterprise. The separate atlas.nomic.ai/pricing app runs Free / Plus $10/mo / Business $125/seat/mo / custom Enterprise, with embedding overage at $1 per 10M text tokens and $1 per 50k images.

AEC Platform pivot — seat + platform-commit model on www.nomic.ai

Nov 2025

Nomic's primary site repositions to a vertical AEC (architecture, engineering, construction) agentic platform. The first archived www.nomic.ai/pricing snapshot (2025-11-15) shows the Nomic Platform Business plan at $40/user/month (annual, 25-seat minimum, $1,000/month platform commitment) with $20 pooled AI usage per seat, plus a custom Enterprise tier. The legacy Atlas tiers and the per-token Embed rate move to atlas.nomic.ai.

Nomic Embed Text v2 — multilingual MoE

Feb 2025

v2 ships as an open multilingual mixture-of-experts embedding model, extending Nomic Embed beyond English while keeping the open-weight, reproducible posture.

Atlas tiered pricing established (Free / Plus / Business / Enterprise)

Oct 2024

The Atlas data-exploration app's tier structure — Free Starter, Plus, Business and custom Enterprise — is in place by the earliest archived atlas.nomic.ai/pricing snapshot (2024-10-06). The page is a JS-rendered SPA, so exact archived dollar figures are not independently legible from the static archive; current figures ($10 Plus, $125/seat Business) are from the 2026-06-03 live capture.

Nomic Embed Vision — multimodal embeddings

Jun 2024

Nomic adds an image-embedding model sharing the text latent space, enabling multimodal search. On Atlas this is metered separately as $1 per 50,000 images after each tier's included image pool.

Nomic Embed v1.5 — Matryoshka resizable embeddings

Feb 2024

v1.5 adds Matryoshka Representation Learning so a single model emits 64-to-768-dimension embeddings, letting users trade vector size against quality at no extra per-token cost.

Nomic Embed text-v1 launches — open-weight embedding model

Feb 2024

Nomic releases nomic-embed-text-v1 (Apache 2.0, 8,192-token context), the first fully reproducible open embedding model to beat OpenAI's text-embedding-ada-002 on MTEB. Embeddings are consumed through Atlas and the Developer API; metered embedding overage on Atlas is quoted at $1 per 10M text tokens.