Vectara pricing

About

Vectara is an enterprise RAG-as-a-Service and agent platform that bills itself as “the enterprise agent platform for trusted enterprise AI: governed, grounded, auditable agents.” It provides a managed retrieval-augmented-generation stack — ingestion, embedding, retrieval, generation, and evaluation — so that enterprises can build grounded conversational AI, customer-support chatbots, and autonomous agents on top of their own data without assembling the underlying vector database, embedding, and LLM-orchestration components themselves.

The platform ships Vectara’s own first-party models — the Boomerang retrieval LLM and the Mockingbird RAG-optimized generative LLM — while also allowing customers to bring their own models such as ChatGPT, Claude, or Gemini. A distinguishing emphasis is governance and trust: Guardian Agents for automated hallucination detection and policy enforcement, a Factual Consistency Score on every answer, workload observability across the RAG pipeline, and automated hallucination correction. Vectara targets regulated and data-sensitive verticals — semiconductors, financial services, eLearning, education, legal, healthcare, telco, and manufacturing.

Vectara is a private company. It was founded as ZIR AI by Cloudera co-founder Amr Awadallah (CEO) with ex-Google engineers Amin Ahmad (CTO) and Tallat Shafaat (chief architect), and emerged from stealth as Vectara in October 2022. It has raised roughly $53.5M — a $28.5M seed led by Race Capital (announced June 2023) and a $25M Series A led by FPV Ventures and Race Capital in July 2024, the latter announced alongside its Mockingbird RAG-tuned LLM.

Rather than seat counts, Vectara’s commercial model is organised around where the platform is deployed — Vectara’s managed cloud (SaaS), a customer’s virtual private cloud (VPC), or fully on-premise — with annual commitments that scale up as the deployment moves closer to the customer’s own environment. This is a striking shift: the company began with a free, self-serve consumption model aimed at developers and moved decisively up-market into six-figure enterprise contracts over 2024–2025. It competes with do-it-yourself RAG stacks and with enterprise-search incumbents such as Glean.

Pricing summary : How Vectara’s deployment-tiered annual commitments work

Vectara uses an annual-commitment model anchored to the deployment tier rather than per-seat or published per-token metering. The public pricing page lists three deployment tiers, each priced as an annual floor and each covering a single deployment:

1. SaaS — starting at $100K/year: 1 SaaS deployment in Vectara’s managed cloud.
2. VPC — starting at $250K/year: 1 VPC deployment into any of the customer’s virtual private clouds; adds Forward-Deployed AI Engineer and Platinum Support.
3. On-prem — starting at $500K/year: 1 fully on-premise deployment; includes Forward-Deployed AI Engineer and Platinum Support.

A 30-day free trial (10,000 credits, all features) replaces the old perpetual free tier. Internally, Vectara meters usage in credits — a unit covering API calls, indexed data volume, and stored data — and the console surfaces queries, storage, and request counts under the Usage and Current plan tabs. The public page does not publish per-credit overage rates; the headline number is the annual commitment floor, with each tier including a base credit allowance.

What makes this different: Vectara prices on where the workload runs — the same platform costs 2.5x more in a customer VPC and 5x more fully on-prem — making the deployment model, not user count or token volume, the primary price axis. This is a commitment-based subscription model with credit-based usage metering layered over agent-platform economics.

Pricing by product

Vectara platform (deployment tiers)

Sales motions across products: PLG / self-serve for the 30-day trial and console signup; sales-led (“Let’s talk”) for all paid SaaS, VPC, and On-prem deployments.

Compare-plans matrix (what each tier includes)

The pricing page’s comparison matrix marks inclusion with checkmarks across all three deployment tiers. Captured verbatim:

In-console plan names

Vectara’s billing documentation references in-console subscription plans named Standard, Pro, Enterprise, VPC, and On-prem available on the console Upgrade tab, with a plan comparison built into the console. The public pricing page surfaces only the deployment-tier view (SaaS / VPC / On-prem).

Hidden costs : What enterprises actually pay beyond the deployment floor

The advertised “starting at” floors understate total cost for larger deployments, because the deployment-size band (Small / Medium / Large) and metered usage (queries, data storage, requests) are negotiated on top of the floor. The floors below are verbatim from the pricing page; the size-band uplift and usage are quoted per deal and are not published.

A mid-size enterprise choosing VPC over SaaS for data residency

The same platform and bundled models cost $150K/year more purely to move the deployment into the customer’s VPC — so data-residency requirements, not feature needs, drive the largest single jump in a Vectara bill. For a deeper look at how deployment choices reshape AI cost, see our blog on AI infrastructure cost drivers.

Want to estimate your own Vectara bill? Use the Vectara pricing calculator to model your annual cost based on deployment tier, size band, and query/storage volume.

Pricing evolution : From a free consumption tier to six-figure deployment commitments

Vectara’s pricing has moved up-market more aggressively than almost any company in this corpus: in roughly three years the entry point went from $0/month (a free Growth tier) to a $100/month subscription to a $100,000/year annual commitment. The shifts below are reconstructed from archived snapshots of the public pricing page.

Cadence

Tracked range: 2022 Q4–2026 Q2. Quarters not listed were verified stable (0 price changes, 0 SKU additions) against archived snapshots.

Notable changes

• 2022-11 — Earliest archived pricing: Growth free tier ($0/mo, 15,000 queries, 50 MB) plus a Scale Contact-Sales tier, with published overage of $1.25/mo per additional 1,000 queries (Wayback snapshot of vectara.com/pricing).
• 2023-06 — Free tier expanded to include 15,000 generative requests/month and Grounded Generation (Wayback snapshot).
• 2024-07 — Vectara raised a $25M Series A (FPV Ventures, Race Capital) and launched its Mockingbird RAG-tuned LLM, bringing total funding to ~$53.5M — the war chest behind the subsequent enterprise repositioning.
• 2024-12 — Pivot to a four-tier subscription page (Standard $100/mo, Pro $6,000/yr, Enterprise, VPC/On-prem) with a new “bundles” usage unit and self-serve credit-card purchase (Wayback snapshot).
• 2025-03 — Visible prices removed from the Standard/Pro/Enterprise cards, signalling a move away from self-serve published pricing (Wayback snapshot).
• 2025-09 — Repriced to six-figure annual deployment tiers ($100K / $250K / $500K) with a 30-day, 10,000-credit free trial; “credits” replaced “bundles” (Wayback snapshot).
• 2025-11 — Deployment tiers renamed SaaS / VPC / On-prem and re-anchored to the same floors, with the “Agent OS for trusted enterprise AI” positioning that remains current (Wayback snapshot).

The up-market migration in detail

The clearest inflection is late 2024 to late 2025, when Vectara abandoned product-led, self-serve consumption pricing for sales-led enterprise commitments. Three archived states bracket the move:

1. Through 2024 Q3 — a free Growth tier (15,000 queries, 50 MB) with $1.25/mo-per-1,000-query overage; the page literally sized its free storage as “enough to load the entire Harry Potter series 5x.”
2. 2024 Q4 — a “Transparent pricing” page with a $100/mo Standard plan and a $6,000/yr Pro plan, billed in “bundles” (1,000 queries + 1,000 Mockingbird generative requests + 10 MB), still self-purchasable by credit card.
3. 2025 Q3–Q4 — six-figure annual deployment tiers ($100K / $250K / $500K), “credits” as the usage unit, and a time-boxed trial replacing the free tier — with every paid tier behind a “Let’s talk” motion.

The change is greater than 5× at every step, but each intermediate state is independently legible in the archive (no interpolation), and the repricing tracks the company’s $53.5M fundraising and its repositioning from a developer-facing neural-search API into an enterprise “trusted AI” agent platform. This mirrors a broader corpus pattern of AI infrastructure vendors retiring free/consumption tiers in favour of committed enterprise revenue.

What’s unique : Deployment-axis pricing and first-party trust models

1. Deployment model is the price axis. Vectara prices by where the platform runs — managed cloud, customer VPC, or on-prem — not by seats or tokens, with each step roughly 2.5x and 5x the SaaS floor.

2. First-party models bundled at every tier. Boomerang (retrieval) and Mockingbird (generative) ship with all tiers, alongside bring-your-own ChatGPT, Claude, or Gemini.

3. Governance built into the product. Guardian Agents, Factual Consistency Score, workload observability, and automated hallucination correction are included across all tiers, not gated behind enterprise add-ons.

Strengths & weaknesses

Billing UX : Console billing tabs, usage metering, and payment controls

• Billing page — shows the current plan and, for trials, the trial end date; it is the entry point for managing payments (per Vectara’s billing docs).
• Usage tab — displays resource usage for the current period, such as the number of requests, plus charges and billing history.
• Current plan tab — shows a plan overview including the amount of data storage and number of queries, plus a Billing details section to edit the billing contact and payment method.
• Upgrade tab — lists available subscription plans (Standard, Pro, Enterprise, VPC, On-prem) and lets you compare plans.
• Add credit card flow — self-serve credit-card entry under Billing details; an “Add credit card” action updates payment information with a confirmation pop-up.
• Billing roles — actions require the Billing Admin or Account Owner role; users without the privilege must request Billing Admin access from the account owner.
• Purchase a plan — trial accounts can select “Purchase a plan” and provide credit-card details to upgrade directly from the console.

Strategic wins : Pricing decisions that worked

1. Publishing real annual floors

Vectara shows concrete numbers ($100K / $250K / $500K) where most enterprise RAG vendors show a blank contact-sales form, which pre-qualifies buyers and signals confidence. See related patterns in our pricing transparency analysis.

2. Pricing on deployment locus, not seats

By making “where the workload runs” the price axis, Vectara aligns price with the cost and risk of VPC and on-prem deployments rather than user count — a clean fit for agent-platform economics where seat counts poorly track value.

3. Bundling the trust suite at every tier

Guardian Agents, Factual Consistency scoring, and hallucination correction ship on all tiers, so governance is a reason to buy rather than an enterprise upsell — reinforcing the “trusted AI” positioning. See our guide to packaging usage-based features.

Areas to improve : Gaps and proposed fixes

1. No self-serve paid entry below six figures

Every paid tier requires “Let’s talk”, leaving the 30-day trial as the only no-touch path. A self-serve console plan (the Standard and Pro tiers already exist in-console) with published pricing would capture smaller teams. See our guide to usage-based pricing.

2. No published usage quotas or overage rates

The public page shows annual floors but never the included query, storage, or request volumes — so buyers cannot self-estimate where they land. Publishing the Small / Medium / Large size bands with concrete query and storage allowances would reduce sales friction. See our blog on usage metering transparency.

3. Support window capped at 5x8 across all tiers

The public comparison matrix shows 5x8 support even on the $500K on-prem tier; a 24x7 option (or making it the default at the top tier) would better match enterprise expectations for a six-figure commitment.

Key takeaways

1. A pricing model can be replaced wholesale, fast. In three years Vectara went from a free consumption tier to $100/mo subscriptions to $100K/yr commitments — proof that an early metering scheme is not a permanent decision if the segment shifts.
2. Deployment model can be the primary price axis. Vectara ladders price by managed-cloud vs VPC vs on-prem rather than seats or tokens, a clean fit when seat counts poorly track value.
3. Publishing floors pre-qualifies buyers. Concrete “starting at” numbers filter out under-budget prospects before sales engages — and quietly signal who the product is now for.
4. Bundling governance avoids nickel-and-diming. Trust features ship at every tier, removing a common enterprise upsell friction.
5. A trial is not a free tier. Vectara’s 30-day, 10,000-credit trial drives evaluation without conceding perpetual free usage — the deliberate replacement for the old Growth free plan.

UBP implications

1. Deployment locus as a value metric. Pricing by where the workload runs is a viable alternative to per-token metering for trust-sensitive AI, and Vectara’s three-tier ladder shows the metric can carry a clean 1x / 2.5x / 5x multiple.
2. The usage unit can evolve with the segment. Vectara moved from raw queries ($1.25/1,000) to “bundles” to “credits” as it climbed up-market — a reminder that the metering abstraction is a packaging lever, not a fixed technical fact.
3. Floors can replace free tiers without losing evaluation. Annual commitment floors anchor revenue while a time-boxed, credit-capped trial preserves a no-cost evaluation path — a template for vendors retiring perpetual free plans.

Sources

• Vectara pricing page (accessed 2026-06-02)
• Vectara console signup (accessed 2026-06-02)
• Vectara billing and manage-payments docs (accessed 2026-06-02)

Bottom line

Vectara prices enterprise RAG and agents by deployment model — SaaS from $100K/yr, VPC from $250K/yr, On-prem from $500K/yr — bundling its own Boomerang and Mockingbird models plus a full governance suite at every tier, with usage metered on queries, storage, and requests behind the annual floor.

Want to compare Vectara against other AI-platform pricing? 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.

Deployment-tiered annual commitments (current)

Jun 2026

Current public pricing: SaaS from $100K/yr, VPC from $250K/yr, On-prem from $500K/yr, each covering a single deployment, with a 30-day all-features free trial and a comparison matrix across sizes, premium add-ons, models, agents, governance, and integration support.

Tiers renamed to SaaS / VPC / On-prem; Agent OS positioning

Nov 2025

Small/Medium/Large were renamed to SaaS ($100K/yr), VPC ($250K/yr), and On-prem ($500K/yr), each with a base credit allowance (10M / 100M / 100M credits/yr) plus purchasable credit packages; messaging shifted to 'Agent OS for trusted enterprise AI.' (Wayback 2025-11-07)

Move up-market: six-figure deployment tiers + credits

Sep 2025

Vectara repriced to enterprise annual deployment tiers — Small starting at $100K/yr, Medium at $250K/yr, Large at $500K/yr — each tied to a deployment (SaaS / VPC / On-prem) and an included credit allowance (100K to 10M credits/yr). 'Credits' replaced 'bundles' as the usage unit, and a 30-day / 10,000-credit free trial replaced the perpetual free tier. (Wayback 2025-09-15)

Headline prices removed from tier cards

Mar 2025

The Standard/Pro/Enterprise cards dropped their visible $100/mo and $6,000/yr price points, replacing them with 'Get started' / 'Request pricing'; the bundles model and self-purchase option stayed. (Wayback 2025-03-25)

Pivot to multi-tier subscription with 'bundles' billing

Dec 2024

Vectara replaced Growth/Scale with a four-card 'Transparent pricing' page: Standard from $100/month (20,000 queries, 200 MB), Pro from $6,000/year, Enterprise (request pricing, 99% SLA), and VPC/On-prem (request pricing). Introduced 'bundles' as the usage unit (1,000 queries + 1,000 Mockingbird generative requests + 10 MB) and self-purchase by credit card. (Wayback 2024-12-02)

Grounded Generation added to free tier

Jun 2023

The Growth free tier gained 15,000 generative requests/month alongside the 15,000 queries, plus Grounded Generation / generative AI summarization; the $1.25/mo per-1,000-query overage and Scale Contact-Sales tier remained. (Wayback 2023-06-09)

Growth (free) + Scale consumption model

Nov 2022

Earliest archived pricing: a two-plan self-serve model — Growth at $0/month (free up to 15,000 queries/month and 50 MB account size) and Scale via Contact Sales. Overage was published at $1.25/month per additional 1,000 queries and 5 MB. (Wayback 2022-11-06)