LlamaIndex

LlamaIndex traces its origins to a side project Jerry Liu began in late 2022. At the time, Liu was a machine learning engineering manager at AI safety startup Robust Intelligence, and previously did self-driving AI research at Uber's Advanced Technologies Group (Uber ATG) and worked on recommendation systems at Quora. While experimenting with OpenAI's GPT-3, Liu ran into its narrow 4,096-token context window and inability to work reliably with private data, so he built a small indexing utility and published it to GitHub in November 2022 under the name 'GPT Tree Index' (later 'GPT Index'). Within a few months the project had accumulated over 16,000 GitHub stars, roughly 200,000 monthly downloads, and a Discord community of about 6,000 developers -- traction that convinced Liu to turn it into a company. He brought on Simon Suo, a former Uber colleague (they had worked together on ML research at Uber) who had since worked at autonomous-driving startup Waabi, as co-founder and CTO. The company was formally incorporated on April 1, 2023, and the project was rebranded LlamaIndex. In June 2023, LlamaIndex announced an $8.5 million seed round led by Greylock Partners, with participation from angel investors Jack Altman, Lenny Rachitsky, and Charles Xie (per TechCrunch and Greylock's own portfolio-news post, 'The Data Framework for LLMs,' dated June 6, 2023 and co-authored by Greylock's Saam Motamedi, Jerry Chen, and Jason Risch). Greylock's post frames LlamaIndex as the 'data framework' layer of an emerging LLM application stack, alongside foundation models, vector databases, and agent frameworks (the post explicitly name-checks LangChain as an agent-framework peer, not yet a head-to-head rival). The seed capital was earmarked to build an 'enterprise solution' on top of the open-source project. Through 2023-2024, LlamaIndex focused on growing the open-source framework (available in Python and TypeScript) and its data-connector ecosystem (LlamaHub), while quietly working with early enterprise users. In March 2025, LlamaIndex announced a $19 million Series A round led by Norwest Venture Partners, with Greylock returning to participate (but not lead) alongside other investors; the round brought total disclosed funding to approximately $27.5 million. Concurrently, LlamaIndex launched LlamaCloud, its first fully commercial, managed cloud offering, and LlamaParse, an 'agentic OCR' product for parsing complex documents (tables, charts, handwriting) using a multi-agent pipeline combining computer vision, specialized vision-language models, and LLM reasoning. At the time of the Series A, CEO Jerry Liu named Salesforce, KPMG, and Carlyle as companies already using LlamaIndex, and said the company's headcount stood at roughly 20 people. That same month, LlamaIndex joined Cisco, LangChain, Glean, and Galileo in launching AGNTCY, an open-source initiative for agent interoperability. In May 2025, LlamaIndex disclosed additional minority equity investments from Databricks (via the Databricks Ventures AI Fund, alongside other Databricks AI bets like Mistral AI, Perplexity, and Cleanlab) and from KPMG LLP (via KPMG Ventures), structured in part around commercial partnerships -- KPMG cited using LlamaCloud/LlamaIndex to build industry-specific AI solutions, and Databricks and LlamaIndex share customers such as Cemex and Carlyle Group. Terms of this investment were not disclosed, so it is excluded from the company's confirmed $27.5M cumulative disclosed funding figure. A December 2024 InfoWorld survey grouped LlamaIndex among the major LLM application frameworks alongside LangChain, Microsoft Semantic Kernel, and Haystack (deepset). Product evolution has been notable: LlamaIndex has progressively repositioned itself from a general-purpose 'RAG framework' toward being a specialized 'agentic document OCR / document intelligence' company. By 2025-2026 its homepage messaging emphasized LlamaParse and 'AI Agents for Document OCR + Workflows' over the original RAG-framework branding, reflecting a bet that unstructured document processing (not general vector search) is the highest-value, most defensible layer of the enterprise AI stack. In October 2025, IBM released an open-source Db2 vector-store connector for LlamaIndex. In March 2026, LlamaIndex released LiteParse, an open-source, offline-capable, TypeScript-native local parsing library (later reworked as a Rust project with TypeScript/Python bindings and markdown output). In April 2026, the company introduced ParseBench, a benchmark for evaluating document-parsing systems, co-authored by CTO Simon Suo among others. As of August 2026, the company's own site claims 25M+ package downloads a month (across its OSS ecosystem), 300k+ LlamaParse users, 1,500+ open-source contributors, and 20,000+ community members; the flagship 'llama-index' PyPI package alone shows roughly 7.6M downloads/month (pypistats.org, via shields.io) and the run-llama/llama_index GitHub repo has roughly 52,000 stars (shields.io badge). Publicly named customers/case studies include Salesforce (Agentforce team), KPMG, The Carlyle Group, Cemex, Jeppesen (a Boeing company), Experian, NTT Data, SkySQL, CiCi AI, 11x.ai, Delphi, GymNation, and others, per the company's own customer-story pages and press materials. Competitively, LlamaIndex is most often compared to LangChain (its closest and most direct peer as an LLM/agent application framework), as well as Microsoft Semantic Kernel and Haystack (deepset) as general-purpose LLM app frameworks; in its newer document-intelligence positioning it also competes with specialized document-parsing/OCR vendors (e.g., Unstructured.io, Reducto, Amazon Textract, Azure AI Document Intelligence, Docugami). LlamaIndex and LangChain have also collaborated at the ecosystem level (both were founding participants in the AGNTCY interoperability initiative). Data-quality caveat: the company's Wikipedia article (accessed August 2026) carries an active 'undisclosed paid editing' maintenance banner, meaning some or all of the article may have been created/edited by paid editors; core facts (founding date, funding amounts/leads, founder identities) were cross-verified against Greylock's own site and TechCrunch reporting and are considered reliable, but softer claims sourced only to Wikipedia should be treated with some caution. No current valuation figure could be confirmed from any public source. No confirmed named board member (beyond the likely Greylock deal lead inference) could be found; specific Greylock board-seat holder is not independently confirmed.

Business model

Open-core / hybrid model: the core LlamaIndex framework (Python and TypeScript) is open source and free, while the commercial LlamaCloud platform (including LlamaParse document parsing/extraction/indexing) is a managed, usage-based SaaS offering sold to individual developers and enterprises, with enterprise-tier contracts (SSO, VPC/hybrid deployment, dedicated support).

Credit-based usage pricing for LlamaCloud/LlamaParse (Free $0/mo with 10K credits; Starter $50/mo; Pro $500/mo; custom Enterprise plans), where 1,000 credits = $1.25; monetization layered on top of the free open-source framework.