LlamaIndex 🦙

@llama_index

Turn any doc into AI-ready context. The AI research lab for document parsing and extraction. Developer-first APIs. LlamaParse: https://cloud.llamaindex.ai/

following
followers

Speculative decoders aim to cut latency for open-source models on OpenDocRouter

LlamaIndex expects considerably faster inference for LightOnOCR-3-1B, but its announcement gives no numerical results.

7 Sources, 1d ago, updated 1d ago

LlamaParse Agentic Plus is said to preserve a research-paper table’s header, counts and labels

LlamaIndex says the “Data Source” header spans Wiki, arXiv, PubMed and SEC, stopping short of “Total.”

2 Sources, 4d ago, first seen 4d ago

Hard document OCR benchmark covers Gemini 4 Argon, Muse Wonder and Muse Spark

The benchmarker calls Muse Spark and Gemini 4 Argon state-of-the-art at their price points for table understanding among frontier models.

5 Sources, 6d ago, updated 6d ago

LlamaIndex uses Markdown by default for parsing, switching to HTML for tables with merged headers

LlamaIndex says Markdown keeps headings, lists and tables intact and stays readable when debugging bad answers.

4 Sources, 11d ago, updated 11d ago

OpenDocRouter launches with 10 document-parsing models under one API

LlamaIndex says users can switch models in one line while keeping the same request and markdown output.

8 Sources, 10d ago, updated 10d ago

Invite-only Agent Economy Gala planned for SF Tech Week on October 9

LlamaIndex says the black-tie Redwood Room event will bring together 120-plus agent founders and 50 investors.

3 Sources, 14d ago, first seen 14d ago

The case for agentic OCR's iterative approach to parsing documents

LlamaIndex contrasts one-pass OCR's trouble with tables and columns with a layout-aware, multi-pass alternative.

2 Sources, 15d ago, first seen 15d ago

LlamaParse reportedly turns BEA's Q2 GDP-by-state infographic into a 52-row table

LlamaIndex says the BEA release puts its key data entirely in an infographic with no text layer.

2 Sources, 18d ago, first seen 18d ago

Daytona held its AI Builders Demo Night at WorkOS’s San Francisco office

Daytona says the monthly event brought AI builders together for demos and conversations, with LlamaIndex partnering and WorkOS hosting.

4 Sources, 20d ago, updated 20d ago

Early approaches to Jev and Jev-like models for document parsing

LlamaIndex says it explored early approaches to Jev and Jev-like models for document tasks including orientation detection, language detection and routing.

3 Sources, 21d ago, updated 21d ago

Form parsing challenges for vision-language models

LlamaIndex argues that forms need purpose-built parsing to preserve section structure, match values to the right boxes and handle handwriting and checkmarks.

3 Sources, 22d ago, first seen 22d ago

LlamaParse reportedly keeps Fed forecast data aligned despite a blank cell

LlamaIndex says all nine numbers in a displayed GDP median excerpt matched the original Fed PDF, while LlamaParse’s HTML output kept the data aligned.

3 Sources, 25d ago, updated 25d ago

LlamaParse card titled “Blank cells need parsing.” An original crop of the Federal Reserve’s September 16, 2026 projections, Table 1 on page 2, shows GDP medians for 2026, 2027, 2028, 2029 and longer run. The current row is 2.3, 2.4, 2.2, 2.1, 2.0 percent. The June row is 2.2, 2.3, 2.2, blank, 2.0. Below it, a red panel marked Incorrect illustrates a hypothetical parsing error: June’s 2.0 is shifted left into the 2029 column and the longer-run cell is empty. A blue panel marked Correct shows the rendered LlamaParse HTML excerpt, with June’s 2029 cell left empty and 2.0 under longer run.

LlamaIndex is building document infrastructure for AI agents

At Connected Stack’s Founder Flash Talks, LlamaIndex highlighted a hurdle for enterprise agents: messy, complex documents that it says general-purpose models struggle to read.

1 Source, 31d ago, first seen 31d ago

‘Parsed by LlamaParse’ series debuts with an energy-report example

LlamaIndex says misreading a table in the EIA’s September 2026 energy outlook can send errors into dashboards, forecasts and AI apps.

1 Source, 32d ago, first seen 32d ago

LlamaIndex announces webinar on insurance document processing

LlamaIndex says the session will use LlamaParse and Extract to turn insurance documents into structured data for underwriting, policy review and claims workflows.

1 Source, 34d ago, first seen 34d ago

LlamaParse’s SDK lessons as the Stainless team reportedly joins Anthropic

LlamaIndex says Stainless saved it SDK maintenance work and pushed it to improve LlamaParse’s API by making inconsistent names and schemas harder to ignore.

1 Source, 35d ago, first seen 35d ago

A two-pass approach to OCR with LiteParse and LlamaParse

LlamaIndex reports a 32-second first pass through a full data room with LiteParse, a free, open-source parser supporting 50-plus formats.

3 Sources, 36d ago, first seen 36d ago

LlamaIndex Finds Higher Cost Does Not Equal Better Extraction

LlamaIndex evaluated 14 frontier systems across 370 enterprise documents and plotted accuracy against cost per page.

1 Source, 46d ago, first seen 46d ago

LlamaIndex Introduces Turbo Mode for Extract

LlamaIndex announces parallel processing mode for its document extraction tool.

6 Sources, 47d ago, first seen 47d ago

LlamaIndex Launches ExtractBench on Kaggle

New benchmark tests schema-guided extraction on complex enterprise documents and scans.

4 Sources, 48d ago, first seen 48d ago

Featured in

Speculative decoders aim to cut latency for open-source models on OpenDocRouter

LlamaParse Agentic Plus is said to preserve a research-paper table’s header, counts and labels

Hard document OCR benchmark covers Gemini 4 Argon, Muse Wonder and Muse Spark

LlamaIndex uses Markdown by default for parsing, switching to HTML for tables with merged headers

OpenDocRouter launches with 10 document-parsing models under one API

Invite-only Agent Economy Gala planned for SF Tech Week on October 9

The case for agentic OCR's iterative approach to parsing documents

LlamaParse reportedly turns BEA's Q2 GDP-by-state infographic into a 52-row table

Daytona held its AI Builders Demo Night at WorkOS’s San Francisco office

Early approaches to Jev and Jev-like models for document parsing

Form parsing challenges for vision-language models

LlamaParse reportedly keeps Fed forecast data aligned despite a blank cell

LlamaIndex is building document infrastructure for AI agents

‘Parsed by LlamaParse’ series debuts with an energy-report example

LlamaIndex announces webinar on insurance document processing

LlamaParse’s SDK lessons as the Stainless team reportedly joins Anthropic

A two-pass approach to OCR with LiteParse and LlamaParse

LlamaIndex Finds Higher Cost Does Not Equal Better Extraction

LlamaIndex Introduces Turbo Mode for Extract

LlamaIndex Launches ExtractBench on Kaggle

Featured in

Speculative decoders aim to cut latency for open-source models on OpenDocRouter

LlamaParse Agentic Plus is said to preserve a research-paper table’s header, counts and labels

Hard document OCR benchmark covers Gemini 4 Argon, Muse Wonder and Muse Spark

LlamaIndex uses Markdown by default for parsing, switching to HTML for tables with merged headers

OpenDocRouter launches with 10 document-parsing models under one API

Invite-only Agent Economy Gala planned for SF Tech Week on October 9

The case for agentic OCR's iterative approach to parsing documents

LlamaParse reportedly turns BEA's Q2 GDP-by-state infographic into a 52-row table

Daytona held its AI Builders Demo Night at WorkOS’s San Francisco office

Early approaches to Jev and Jev-like models for document parsing

Form parsing challenges for vision-language models

LlamaParse reportedly keeps Fed forecast data aligned despite a blank cell

LlamaIndex is building document infrastructure for AI agents

‘Parsed by LlamaParse’ series debuts with an energy-report example

LlamaIndex announces webinar on insurance document processing

LlamaParse’s SDK lessons as the Stainless team reportedly joins Anthropic

A two-pass approach to OCR with LiteParse and LlamaParse

LlamaIndex Finds Higher Cost Does Not Equal Better Extraction

LlamaIndex Introduces Turbo Mode for Extract

LlamaIndex Launches ExtractBench on Kaggle