LlamaIndex pricing

About

LlamaIndex is the company behind one of the most widely used open-source RAG and agent orchestration frameworks — started as “GPT Index” by Jerry Liu in November 2022 and renamed in February 2023 — and a hosted platform (LlamaCloud, now fronted by LlamaParse) that turns messy documents into agent-ready data: parsing, extraction, classification, splitting, spreadsheets, and managed indexes. Founded in 2023 by Jerry Liu and Simon Suo in San Francisco, it raised an $8.5M seed led by Greylock (June 2023) and a $19M Series A led by Norwest Venture Partners (March 2025), bringing total funding to $27.5M. By the Series A it reported more than 3 million monthly downloads, nearly 40,000 GitHub stars, and customers including Rakuten, Carlyle, Salesforce, and KPMG.

The monetization split is clean: the framework is free open source; the money is in document processing, billed in credits where each action — parsing a page, extracting structured data, indexing — burns a posted credit amount, and 1,000 credits cost $1.25.

For the most current information, visit LlamaIndex.

Pricing summary : How LlamaIndex’s pricing model works

LlamaIndex’s platform bills in a single currency — credits, at $1.25 per 1,000 (identical in the US and EU regions) — across three self-serve tiers plus a quoted Enterprise plan. Free ($0) includes 10K credits/month for 1 user with no pay-as-you-go. Starter ($50/month) includes 40K credits, 5 users, and pay-as-you-go up to 400K credits (a $500/month cap). Pro ($500/month) includes 400K credits, 10 users, higher concurrency (20 parallel parse jobs vs 5), and pay-as-you-go up to 4,000K credits ($5,000/month cap). Enterprise is custom — volume discounts on credits, 5x rate limits, SSO, SaaS or hybrid-cloud (VPC) deployment, and a dedicated account manager.

What a credit buys depends on the job. Parsing runs 1 credit/page (Fast), 3 (Cost-effective), 10 (Agentic), or 45 (Agentic Plus); extraction adds 5–15 credits per page on top; layout detection adds 3; classification costs 1–2; indexing 1–2; audio is 3 credits per minute. Re-parsing a cached file is free.

What makes this different: the subscription is arithmetically a prepaid credit bundle — Starter’s $50 equals exactly 40,000 credits at $1.25/1K, and Pro’s $500 equals 400,000. There is no bundled-credit discount; what tiers actually sell is seats, concurrency, index capacity, and support. That design made it possible for LlamaIndex to raise the per-credit price 25% in February 2026 while trimming included credits, without the $50/$500 stickers ever moving (see Pricing evolution).

Pricing by product

Sales motions across products: self-serve PLG for Free/Starter/Pro (instant signup, card checkout), open-source adoption as top-of-funnel, and sales-led for Enterprise. A startup program offers free credits and dedicated support by application.

Hidden costs : What LlamaIndex users actually pay

The plan price is just a prepaid floor — the real bill is set by which tier you parse at. The same page costs 1 credit on Fast ($0.00125), 3 on Cost-effective ($0.00375), 10 on Agentic ($0.0125), and 45 on Agentic Plus ($0.05625). Add extraction (5–15 credits/page on top of parse), layout detection (+3/page), or audio (3 credits/minute) and a “parse plus extract” pipeline runs 6–60 credits per page — a 10x-plus spread driven entirely by configuration choices.

Watch the pay-as-you-go caps: Starter overage stops at 400K credits ($500/month) — past that you must jump to the $500 Pro plan, a 10x sticker jump. Heavier extraction stacks (Agentic extract on Agentic Plus parse, 60 credits/page) can burn Starter’s entire included allotment in about 670 pages. The cushions: smart caching makes re-parsing a previously parsed file free, Auto Mode routes each page to the cheapest adequate tier (LlamaIndex claims savings up to 80%), and classification (1–2 credits/page) lets you pre-filter before expensive parses.

Want to estimate your own LlamaIndex bill? Use the LlamaIndex pricing calculator to model your costs based on usage patterns.

Pricing evolution : LlamaIndex pricing history and changes

Cadence

Tracked range: 2024–present, via LlamaIndex blog announcements and Wayback Machine snapshots (2025-05 through 2026-04) plus a live 2026-06-10 capture.

Notable changes

• 2024-03 — LlamaParse paid launch: 1,000 pages/day free (plus 7,000 pages/week), then $0.003/page ($3 per 1,000 pages). Documents capped at 750 pages.
• 2025-03 — $19M Series A (Norwest) announced alongside platform general availability as SaaS and on-premise, with SOC 2 Type 2 and an EU GDPR-compliant option; free and paid plans with 10K free signup credits.
• 2025-05 (earliest archived grid) — Credit-based plans: Free 10K credits, Starter $50 with 50K credits (pay-as-you-go to 500K), Pro $500 with 500K credits (pay-as-you-go to 5,000K). 1,000 credits = $1.00.
• 2025-12 — LlamaParse v2 replaces dozens of mode/model combos (1–90 credits/page) with four tiers: Fast (1), Cost-effective (3), Agentic (10), Agentic Plus (45) — the top tier 50% cheaper than v1 agent modes.
• 2026-02 — Credit repriced to $1.25 per 1,000 (+25%); included credits trimmed to 40K (Starter) and 400K (Pro), pay-as-you-go caps reduced in step. Headline $50/$500 prices unchanged — the increase is invisible on the sticker.

What’s unique : LlamaIndex’s distinctive pricing mechanics

1. The subscription is a prepaid credit bundle at face value. $50 buys exactly $50 of credits; $500 buys exactly $500. There’s no volume discount inside the self-serve tiers (discounts start at Enterprise), so tier upgrades are purely about seats, concurrency, indexes, and compliance. That’s unusually honest arithmetic — and it makes the meter, not the plan, the real pricing surface.

2. One meter, sixty burn rates. Every product — parse, extract, classify, split, index, audio, spreadsheets — burns the same credit currency at posted per-page rates spanning 1 to 60 credits. That lets LlamaIndex ship new SKUs (Sheets, Agents are currently free in beta) without new pricing pages: a new feature is just a new burn rate on the existing meter.

3. Repricing the credit, not the plan. The February 2026 change moved the per-credit price 25% and quietly shrank included allotments about 20%, while the $50/$500 anchors never moved. Customers comparing screenshots month-over-month would see identical plans — only the fine print (“Includes 40K credits”, “1,000 credits = $1.25”) changed.

Strengths & weaknesses

Billing UX : LlamaIndex billing controls and transparency

• Billing controls — Self-serve signup and card checkout for Free/Starter/Pro with a region picker (US or Europe) at sign-up; pay-as-you-go is capped per plan (Starter $500/mo, Pro $5,000/mo), which doubles as a built-in spend ceiling. Enterprise is invoiced with volume discounts.
• Usage visibility — The docs publish the full credit schedule per tier and action, and the dashboard tracks credit consumption per job. Cached re-parses cost 0 credits and Auto Mode picks the cheapest adequate parse tier per page — both legible cost-control levers.
• Payment options — Card for self-serve; sales-negotiated contracts (SaaS or hybrid VPC deployment, custom BAAs) for Enterprise. A startup program offers free credits and dedicated support by application.

Strategic wins : Why LlamaIndex’s pricing decisions worked

1. Monetizing the pipeline, not the framework

LlamaIndex never tried to charge for the open-source framework that made it famous (3M+ monthly downloads by 2025). Instead it metered the adjacent pain — turning PDFs, decks, and spreadsheets into agent-ready data — where value per page is concrete and willingness to pay is enterprise-grade (Rakuten, Carlyle, Salesforce, KPMG). See how AI companies structure pricing.

2. A single credit currency across every SKU

Parse, extract, classify, split, index, and audio all burn one credit meter with posted rates. New products launch as free betas (Sheets, Agents) and graduate to a burn rate — no repricing event, no new plan grid. That’s a textbook pattern for multi-product usage pricing; see choosing the right usage metric.

3. Simplifying the meter when complexity hurt conversion

v1 parsing exposed dozens of mode/model combinations costing 1–90 credits per page, and users reported the choice burden. The December 2025 v2 collapse to four named tiers — with the premium tier cut 50% — traded configurability for forecastability, plus Auto Mode to automate the tradeoff. Related: outcome-based pricing trends.

Areas to improve : Gaps in LlamaIndex’s pricing approach

1. Announce repricings instead of editing fine print

The February 2026 change — credit price up 25%, included credits down about 20% — shipped with no prominent announcement we could find; only the pricing page’s fine print changed. For a developer audience that screenshots pricing pages, silent repricing erodes trust faster than the increase itself. See bill shock and cost unpredictability.

2. Smooth the Starter-to-Pro cliff

Starter’s pay-as-you-go hard-stops at 400K credits ($500/month), and the next rung is a $500/month plan — a 10x sticker jump with nothing in between. A mid-tier or an uncapped pay-as-you-go rate would keep growing teams from hitting an artificial wall just as their usage proves out.

3. Reconcile the pricing page with itself

On the live page, the Starter card says 5 users while the compare table says 10; Pro’s card says 10 while the table says 25; the Free card promises “Basic support” while the table lists “Community.” For a company whose credit schedule is admirably precise, the plan-grid inconsistencies undermine an otherwise transparent page.

Key takeaways

1. Sell the meter, not the bundle. LlamaIndex’s plans are prepaid credit packs at face value — the real pricing decisions live in the per-action credit schedule, which is where buyers should look too.
2. A single credit currency scales across products. Six-plus SKUs share one meter with posted burn rates, letting new features launch without new pricing machinery.
3. Credit systems enable invisible repricing. Moving the per-credit price and the included allotment — while keeping sticker prices fixed — delivered a 25% unit increase with zero headline change.
4. Configuration is the real cost driver. The same page costs 1–60 credits depending on tier and extraction choices; tooling like Auto Mode and free cached re-parses matter more than plan choice.
5. Open source funds the funnel, documents fund the business. The framework stays free and ubiquitous; monetization happens where unstructured data meets enterprise budgets.

UBP implications

1. Credit-denominated pricing decouples price from packaging. Vendors gain a lever to reprice units without touching plans — buyers should track the credit value and included allotments, not the sticker. See usage-based pricing strategy.
2. Posted burn-rate schedules build trust, but only with stability. Publishing exact credits-per-action is a transparency win; changing the credit’s dollar value quietly spends that trust down.
3. Tier-based meters beat config-based meters. v2’s four named tiers (1/3/10/45 credits) replaced a 1–90 credit configuration space — evidence that buyers prefer a small set of forecastable rates over fine-grained control.

Sources

• LlamaIndex pricing page — live capture (accessed 2026-06-10)
• LlamaParse credit pricing docs — live capture (accessed 2026-06-10)
• Launching the first GenAI-native document parsing platform — published 2024-03-13 (accessed 2026-06-10)
• Announcing our Series A and LlamaCloud general availability — published 2025-03-04 (accessed 2026-06-10)
• Introducing LlamaParse v2: simpler, better & cheaper — published 2025-12-18 (accessed 2026-06-10)
• LlamaIndex raises $8.5M seed (BusinessWire) — June 2023 (accessed 2026-06-10)
• LlamaIndex secures $19M Series A (PR Newswire) — March 2025 (accessed 2026-06-10)
• LlamaIndex GitHub repository (run-llama/llama_index) (accessed 2026-06-10)
• Wayback Machine snapshots: llamaindex.ai/pricing — 2025-05, 2025-09, 2025-11, 2026-01, 2026-02, 2026-04 (accessed 2026-06-10)

Bottom line

LlamaIndex pairs a free, ubiquitous open-source RAG/agent framework with a credit-metered document platform: Free (10K credits/month), Starter $50 (40K credits), Pro $500 (400K credits), and quoted Enterprise, with 1,000 credits at $1.25 and pay-as-you-go overage capped per plan. Every plan’s price equals its included-credit value, so the per-action credit schedule — 1 to 60 credits per page depending on parse and extract tier — is the real pricing surface. The February 2026 repricing (credit +25%, allotments −20%, stickers unchanged) shows both the power and the trust risk of credit-denominated pricing. Browse the pricing blueprint for more fully-researched company profiles.

Want to compare LlamaIndex against other document and data-pipeline companies like Firecrawl, Exa, or Vectara? 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.

Current: prepaid credit bundles at $1.25 per 1K

Jun 2026

Free ($0, 10K credits), Starter ($50/mo, 40K credits + pay-as-you-go to $500), Pro ($500/mo, 400K credits + pay-as-you-go to $5,000), Enterprise custom with volume discounts and 5x rate limits. US and EU regions at price parity.

Credit repriced $1.00 → $1.25; included credits trimmed

Feb 2026

Credit price rises 25% to $1.25 per 1,000 while included credits drop about 20% (Starter 50K→40K, Pro 500K→400K; pay-as-you-go caps 500K→400K and 5,000K→4,000K). Headline plan prices unchanged at $50/$500 — plan price now exactly equals included-credit value.

LlamaParse v2 — four parse tiers replace modes

Dec 2025

v2 replaces the mode-and-model maze (1–90 credits/page) with four tiers: Fast (1 credit), Cost-effective (3), Agentic (10), Agentic Plus (45) — the top tier a 50% price cut versus v1 agent modes.

Credit-based Free / Starter / Pro grid (earliest archived)

May 2025

Earliest archived view of the credit-plan grid: Free ($0, 10K credits), Starter ($50/mo, 50K credits, pay-as-you-go to 500K), Pro ($500/mo, 500K credits, pay-as-you-go to 5,000K), Enterprise custom. 1,000 credits = $1.00.

$19M Series A + LlamaCloud general availability

Mar 2025

LlamaIndex announces its Norwest-led $19M Series A (total raised $27.5M) alongside platform general availability as SaaS and on-prem, with free and paid plans and 10K free signup credits.

LlamaParse paid launch — per-page pricing

Mar 2024

Three weeks after LlamaParse's public launch, paid plans arrive: 1,000 pages/day free (plus 7,000 pages/week), then $0.003 per page — $3 per 1,000 pages. Documents capped at 750 pages.