Weaviate pricing

About

Weaviate is an open-source, AI-native vector database. The company was founded in 2019 in Amsterdam by Bob van Luijt, open-sourced the core engine under the permissive BSD-3 license, and built a commercial managed service (Weaviate Cloud) on top. In April 2023 it raised a $50M Series B led by Index Ventures (with Battery Ventures, NEA and others; roughly $68M raised in total), and by 2023–2024 reported 90+ employees, 15M+ downloads and a 4,000+ member community.

Its positioning is the classic open-core flywheel: the same database powers a free laptop install, a transparent pay-as-you-go SaaS, and negotiated enterprise deployments. Pricing is its signature — Weaviate bills the managed service on vector dimensions stored, a metric founder Bob van Luijt has publicly argued is the most honest, predictable unit for a vector DB.

As of August 2026, Weaviate ships a second, separately-billed product: Engram, managed memory infrastructure for AI agents, now generally available and priced per pipeline run rather than per vector dimension (see Pricing by product).

For the most current information, visit Weaviate.

Pricing summary : How Weaviate’s pricing model works

Weaviate prices its managed Weaviate Cloud service on two usage dimensions — vector dimensions stored (per 1M dimensions/month) and storage (per GiB) — then wraps that usage in three named plan tiers that set the SLA, support response, backup retention and feature ceiling (SSO, dedicated deployment). The plans are Free ($0, 1 cluster per user), Flex (starts at $45/mo, pay-as-you-go) and Premium (starts at $400/mo, prepaid contract, shared or dedicated deployment).

On top of any tier sit usage-based AI Services (included on every plan, billed by usage): Weaviate Embeddings (per million tokens) and the agentic Query Agent (per request, free up to 250 ask-queries or 1,000 search-queries/month). The full database is also free and open source under BSD-3 to self-host, and Dedicated Cloud / Enterprise Cloud / BYOC deployments (HIPAA on AWS) are sales-led and quoted. For background on this style of metering, see our introduction to usage-based pricing.

Weaviate also runs Engram, a separately-billed managed-memory product for AI agents that reached general availability in August 2026. Engram is priced on its own pipeline run unit (one POST /memories call = one run) across a Free/Starter/Team/Enterprise ladder — a distinct billing dimension from the database’s per-vector-dimension model. See the Engram sub-section under Pricing by product for its own plan grid and overage rates.

What makes this different: Weaviate bills the database on vector dimensions stored rather than nodes, pods or RAM. The headline plan floors ($45 / $400) are low; real cost is driven by how many dimensions you store (objects × embedding dimension size) and how much disk they consume — so the plan tier mostly buys you a better SLA, support, SSO and a dedicated-deployment option, not a different unit rate. Each step up actually lowers the per-dimension rate: $0.00465/1M on Flex → $0.003875/1M on Premium shared → $0.002718/1M on Premium dedicated.

Pricing by product

Engram (managed AI agent memory) — separately billed

Engram is Weaviate’s managed-memory product for AI agents, generally available as of August 2026. It bills per pipeline run — each POST /memories call is one run, regardless of conversation length or payload size — on its own plan ladder, independent of the Database product’s per-dimension pricing.

Sales motions across products: self-serve PLG for Free/Flex/Engram Free-Starter-Team (sign up, pay by card), prepaid contracts for Premium, and sales-led for Dedicated/Enterprise/BYOC/Engram Enterprise. Cloud-marketplace billing is available via AWS and GCP.

Hidden costs : What Weaviate users actually pay

The plan floor is rarely the real bill. Cost is driven by dimensions stored = objects × embedding dimension size, plus disk. A 1M-object index using a 1,536-dimension OpenAI embedding stores ~1.536B dimensions; at Flex’s $0.00465/1M that’s ~$7,140/month before storage, replication or the $45 floor — so naïve “it starts at $45” estimates can be off by orders of magnitude for production indexes. Below is an illustrative mid-size Flex workload.

Other cost levers to watch: Premium is a prepaid contract (you commit up front), high-availability/multi-AZ replication multiplies stored dimensions, dedicated Premium deployments carry a higher floor than the shared-cluster rate suggests, and data transfer is currently free only as a promotional period — a future re-introduction would be a real cost change.

Want to estimate your own Weaviate bill? Use the Weaviate pricing calculator to model dimensions stored, storage and embedding tokens against the Flex / Premium rates.

Pricing evolution : Weaviate pricing history and changes

Cadence

Tracked range: 2023–present, via Wayback Machine snapshots of weaviate.io/pricing (Jan 2023, Jan 2024, Jan 2025, Jan 2026).

Notable changes

• 2023-01 — Weaviate Cloud Services launches per-dimension PAYG pricing at $0.05 per 1M vector dimensions with a free Sandbox and three SLA tiers (Standard / Enterprise / Business Critical).
• 2024-01 — Serverless Cloud GA with a published rate card: Standard $25/mo ($0.095/1M), Enterprise $135/mo ($0.145/1M), Business Critical $450/mo ($0.175/1M) — per-dimension rates roughly doubled as the product commercialized.
• 2025-01 — Weaviate Embeddings (from $0.04/M tokens) and the agentic Query Agent added as usage-based add-ons; pricing page reorganized around deployment options.
• 2026-01 — Major restructure: named plans Free / Flex ($45) / Plus ($280, new) / Premium ($400) replace the SLA-tier naming; the entry per-dimension rate falls ~95% to $0.00465/1M; annual prepaid commitments introduced on Plus and Premium.
• 2026-07-06 — Plus tier removed; ladder simplified to Free / Flex / Premium. The $280/mo mid-tier (SSO, 99.9% uptime) was dropped and its differentiators folded into Premium, which now carries SSO/SAML, a 99.95% uptime SLA and a shared-or-dedicated deployment choice (4-hour Sev-1 on shared, 1-hour on dedicated). Per-dimension rates were left unchanged ($0.00465 / $0.003875 / $0.002718 per 1M for Flex / Premium-shared / Premium-dedicated) — so this is a packaging simplification, not a repricing. The practical effect: the only monthly, no-commit tier is now Flex, and any SSO or tighter-SLA requirement pushes a buyer straight to a prepaid Premium contract.
• 2026-08-28 — Engram launches: a separately-billed managed-memory product for AI agents, GA on its own Free / $45 Starter / $360 Team / custom Enterprise ladder priced per pipeline run (Free 1,000 runs; Starter 10,000 runs + $0.0045/overage; Team 90,000 runs + $0.0035/overage; Enterprise custom volume). Discovered via a “Database” vs “Engram” product-family tab at the top of weaviate.io/pricing, this is a new product line and a new billing unit — not a change to any existing tier. Weaviate Cloud database pricing (Free / Flex $45 / Premium $400, per-dimension rates) was verified unchanged.

What’s unique : Weaviate’s distinctive pricing mechanics

1. Vector dimensions as the billing unit. Where most managed vector DBs bill on pods, nodes, read/write units or RAM, Weaviate bills on dimensions stored (objects × embedding size). Founder Bob van Luijt has defended this publicly as the metric a buyer can actually compute up front — embedding dimension, objects, queries — making the bill predictable from first principles.

2. Plan tiers buy SLA, not (mostly) unit rate. The price step from Flex → Premium mostly purchases uptime SLA (99.5% → 99.95%), Sev-1 response (next-business-day → 4-hour on shared, 1-hour on dedicated), backups (7 → 30 → 45 days), SSO/SAML and the dedicated-deployment option with a Technical Account Team — not a cheaper raw dimension rate alone. Usage rate and reliability are decoupled: each step up actually lowers the per-dimension rate ($0.00465/1M Flex → $0.003875/1M Premium shared → $0.002718/1M Premium dedicated).

3. Three commercial models off one BSD-3 engine. Free self-hosted OSS, transparent self-serve PAYG SaaS, and sales-led Enterprise/Dedicated/BYOC all run the identical database — so pricing scales with commitment (free → monthly card → annual prepaid → quoted contract) without a product migration.

4. A second product gets a second, purpose-built billing unit — not a shared meter. When Engram (managed AI-agent memory) reached general availability in August 2026, Weaviate didn’t stretch the database’s per-dimension metric to cover it. Engram bills per pipeline run (one POST /memories call, regardless of conversation length or payload size), on its own Free/Starter/Team/Enterprise ladder with its own overage rate. That’s the same “buyer can compute the metric themselves” discipline from point 1, reapplied to a workload dimensions can’t describe — a deliberate choice over the easier path of bolting a discount code onto an existing plan.

Strengths & weaknesses

Billing UX : Weaviate billing controls and transparency

• Billing controls — Self-serve sign-up and credit-card billing on Free/Flex via the Weaviate Cloud Console; Premium is a prepaid commitment contract (“Contact Sales” to start); cloud-marketplace billing available through AWS and GCP for consolidated invoicing. Weaviate Cloud runs billing on Stripe.
• Usage visibility — Pricing is computed from vector dimensions stored, data objects and queries; the pricing page ships an interactive Pricing Calculator (Flex vs Premium, object count, object size, Cost-Optimized vs Performance-Optimized index profile) that returns an estimated $/mo, and the Console exposes a Running Costs module plus Console metrics on every plan (metrics endpoint on Premium).
• Payment options — Credit card (self-serve on Free/Flex), prepaid-commitment invoicing (Premium), and cloud-provider marketplace billing (AWS, GCP). Dedicated/Enterprise/BYOC are invoiced via sales contracts.
• Engram overage metering — Starter and Team plans are never rate-limited: pipeline runs beyond the monthly allowance are metered and billed in arrears at the start of the next cycle ($0.0045/run on Starter, $0.0035/run on Team), alongside that period’s base fee. The Free plan instead throttles at the cap with no overage billing.

Strategic wins : Why Weaviate’s pricing decisions worked

1. Picking a metric buyers can compute themselves

By billing on vector dimensions stored, Weaviate gave buyers a cost they can predict before signing up — embedding size × objects × queries. That transparency is a trust asset in a category prone to opaque pod/RU pricing. See choosing the right usage metric.

2. Free OSS + free sandbox as the top of funnel

A BSD-3 free engine plus a no-credit-card managed sandbox lets developers prove value before paying, then graduate to Flex and beyond. This open-core flywheel is the engine behind 15M+ downloads. Related: how AI companies structure pricing.

3. Passing unit-cost deflation to customers

Cutting the entry per-dimension rate ~95% between 2024 and 2026 tracks the broader collapse in compute and embedding costs — and converts deflation into a competitive weapon against Pinecone, Qdrant and pgvector rather than pocketing it as margin. See token cost deflation.

4. Launching a new product on its own meter, not a bolt-on

By billing Engram per pipeline run instead of retrofitting the per-dimension database metric, Weaviate avoided a confusing price translation — “how many dimensions is a memory retrieval?” — and instead gave agent-builders a unit tied directly to how they call the API. It’s the same computable-metric discipline as the core database, applied to open a second product line (GA August 2026) without diluting the original pricing story. See choosing the right usage metric.

Areas to improve : Gaps in Weaviate’s pricing approach

1. The floor-vs-reality gap

“Starts at $45/mo” undersells the true cost of production indexes, where stored dimensions dominate. Clearer “typical bill at X objects” worked examples on the pricing page would reduce bill shock for first-time buyers.

2. The Flex-to-Premium cliff after removing Plus

With the mid-tier Plus removed in mid-2026, any team that needs SSO/SAML, a tighter SLA or a dedicated deployment now jumps straight from month-to-month Flex to a prepaid Premium contract — there is no self-serve monthly step in between. A monthly Premium-lite (or an SSO add-on on Flex) would smooth that cliff for security-conscious teams not yet ready to commit up front.

3. Promotional “free data transfer”

Flagging data transfer as free for now is honest but leaves a known unknown in the bill. Committing to a permanent egress policy (or publishing the future rate) would remove a lingering predictability gap.

4. No cross-product bundling for customers running both Engram and the database

A team using Weaviate Cloud for storage and Engram for agent memory pays two separate bills against two separate ladders, with no combined-usage discount or consolidated invoice advertised on the pricing page. A bundled “Database + Engram” tier — or at minimum a single usage dashboard spanning both meters — would cut the operational overhead for teams that are the obvious target buyer for both products at once.

Key takeaways

1. Bill on what the buyer can compute. Dimensions stored is Weaviate’s defining choice — predictable, transparent, and defensible in public by the founder — and the same discipline carried over to Engram’s per-pipeline-run metric when the company launched a second product in August 2026.
2. Decouple reliability from unit rate. The Flex → Premium upgrade buys SLA, support, backups and SSO — and actually lowers the per-dimension rate rather than raising it. Reliability is a separate axis from consumption.
3. Open core is the funnel. Free BSD-3 self-host + no-card sandbox feed paid Cloud — the same engine, scaled by commitment level.
4. Deflation is a strategy. A ~95% per-dimension price cut over two years is a deliberate competitive lever, not just cost pass-through.
5. Watch the floors. Low monthly minimums ($45 Flex / $400 Premium) make Weaviate look cheap; storage-heavy indexes are where the real spend lives.

UBP implications

1. Choose a value metric the customer already understands. Weaviate’s per-dimension model shows that a unit tied to the buyer’s own data (objects × embedding size) beats infrastructure-shaped units for trust and predictability — and the company didn’t force that same metric onto a differently-shaped product: Engram’s August 2026 launch introduced pipeline runs as its own unit rather than stretching “dimensions” to cover agent memory.
2. Package usage inside SLA tiers. Wrapping pure usage in Free/Flex/Premium lets Weaviate monetize reliability and support separately from raw consumption — a clean pattern for any usage-based product.
3. Use price cuts offensively. Passing unit-cost deflation straight to the rate card (95% over two years) turns falling COGS into a pricing weapon — relevant to every team in the Infrastructure, Compute & MLOps space facing commoditizing compute.

Sources

• Weaviate pricing page (accessed 2026-07-06)
• Weaviate Shared Cloud deployment (accessed 2026-07-06)
• Weaviate Dedicated Cloud deployment (accessed 2026-07-06)
• Weaviate open-source database (GitHub, BSD-3) (accessed 2026-07-06)
• Weaviate Query Agent docs — usage & free-tier limits (accessed 2026-08-28) — independently confirms the free tier is “up to 250 ask queries or 1000 search queries per month,” used to correct the pricing page’s flattened “1,000 requests/month” figure
• Weaviate Cloud billing docs (accessed 2026-07-06)
• Wayback Machine snapshots of weaviate.io/pricing: Jan 2023, Jan 2024, Jan 2025 (accessed 2026-06-09)
• “Thoughts on Pricing a Vector Database” — Bob van Luijt, Hacker News, July 2023 (accessed 2026-06-09)
• Weaviate Series B funding announcement, Index Ventures, April 2023 (accessed 2026-06-09)

Bottom line

Weaviate prices its managed vector database on a metric the buyer can compute — vector dimensions stored — and wraps it in a Free → Flex → Premium ladder where the upgrade buys SLA, support and SSO while actually lowering the per-dimension unit rate. A free BSD-3 self-host option and a no-credit-card sandbox feed the funnel, while a ~95% per-dimension price cut over two years turns falling compute costs into a competitive weapon. The catch: low plan floors hide the fact that large indexes are billed on stored dimensions, so model your real object count before assuming “$45/mo.” As of August 2026, that same buyer-computable-metric discipline extends to a second product — Engram, managed AI-agent memory billed per pipeline run — run as its own ladder rather than folded into the database’s per-dimension price. Browse the pricing blueprint for more fully-researched company profiles.

Want to compare Weaviate against other vector databases and Infrastructure, Compute & MLOps companies? Browse 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.

Engram launches — separately-billed managed memory for AI agents

Aug 2026

Weaviate GA'd Engram, a new managed-memory product for AI agents billed per pipeline run on its own Free ($0, 1,000 runs)/Starter ($45/mo + $0.0045 overage, 10,000 runs)/Team ($360/mo + $0.0035 overage, 90,000 runs)/Enterprise (custom) ladder — a new product line and billing unit, not a repricing. Weaviate Cloud database pricing (Free/Flex $45/Premium $400, per-dimension rates) was verified unchanged.

Plus tier removed — plans simplified to Free / Flex / Premium

Jul 2026

The Plus tier ($280/mo, SSO, 99.9%) was dropped; the plan ladder is now Free, Flex ($45/mo PAYG) and Premium ($400/mo prepaid contract). SSO/SAML, 99.95% uptime and the shared-or-dedicated deployment choice all consolidate into Premium; Premium shared Sev-1 response is 4 hours (dedicated 1 hour). Per-dimension rates unchanged ($0.00465/$0.003875/$0.002718 per 1M for Flex/Premium-shared/Premium-dedicated).

Plan restructure: Free / Flex / Plus / Premium

Jan 2026

SLA-tier naming (Standard/Enterprise/Business Critical, Sandbox) replaced by named plans Free, Flex ($45/mo PAYG), Plus ($280/mo annual prepaid, new), and Premium ($400/mo annual prepaid). Per-dimension rate dropped ~95% (from ~$0.095 to $0.00465 per 1M on the entry paid tier); annual prepaid commitments introduced for the upper tiers.

AI Services bundled in; Weaviate Embeddings priced per token

Jan 2025

Serverless Cloud foregrounded as the primary deployment, with Weaviate Embeddings introduced as a usage-based add-on starting at $0.04/M tokens and the agentic Query Agent added. SLA tiers retained; pricing page reorganized around deployment options (Serverless / Dedicated / self-host).

Serverless Cloud GA with published per-tier rate card

Jan 2024

Serverless Cloud goes generally available with explicit SLA-tier pricing: Standard from $25/mo ($0.095 per 1M dimensions/mo), Enterprise from $135/mo ($0.145/1M), Business Critical from $450/mo ($0.175/1M). Per-dimension rates roughly doubled vs the 2023 beta as the model commercialized.

Per-dimension Serverless model in private beta

Jan 2023

Weaviate Cloud Services launches pay-as-you-go pricing 'per vector dimension stored and queried' starting at $0.05 per 1M vector dimensions, across three SLA tiers (Standard, Enterprise, Business Critical) with a free Sandbox. The per-dimension metric becomes the company's signature pricing choice.