Skip to content

LlamaIndex · Document API

LlamaParse Extract Agentic on ResumeBench: 85.5 F1, rank 3 of 18

LlamaParse Extract Agentic ranks 3rd of 18 on ResumeBench: 85.5 F1, $23 per 1,000 pages, 55 seconds per resume. It is 1.6 points behind LlamaParse Extract Agentic Plus.

3rd of 7 document APIs. Strongest on handwritten (+12.0 points), image only (+10.1 points) and scanned & rotated (+8.7 points). Weakest on medium (−0.2 points).

Rank
3/18
F1
85.5
Precision
86.8
Recall
87.3
Perfect resumes
41/300
Per 1,000 pages
$23
Median per resume
55 s

Averaged over 300 resumes. Dates (scored separately): 97.7. 90% finish within 77 s.

Where LlamaParse Extract Agentic wins and loses

F1 by kind of resume, against the median system.

Reading the page

LlamaParse Extract Agentic by reading the page
Kind of resume# of resumesF1MedianGapRank
Handwritten963.551.5+12.02nd
Image only2881.571.4+10.12nd
Sidebar or grid8285.581.9+3.52nd
Photo4584.784.0+0.68th

Distractors

LlamaParse Extract Agentic by distractors
Kind of resume# of resumesF1MedianGapRank
Rotated & warped pages2190.982.9+8.02nd
Hidden text & prompt injections1491.684.5+7.02nd
Content as images891.688.2+3.44th
Wording traps1687.784.5+3.23rd
Typos kept on purpose3790.083.7+6.42nd

Extraction challenges

LlamaParse Extract Agentic by extraction challenges
Kind of resume# of resumesF1MedianGapRank
Many entries8383.081.9+1.28th
Long bullet lists10183.181.4+1.78th
Split across pages8583.882.2+1.67th
Big skills list6083.783.6+0.19th
Custom sections10582.580.8+1.75th
Combined headings3783.282.7+0.57th
Promotions8885.583.5+2.05th
Partial dates13885.683.4+2.13rd
Overlapping jobs6983.581.7+1.85th
Accented text1391.185.5+5.62nd

Noise

LlamaParse Extract Agentic by noise
Kind of resume# of resumesF1MedianGapRank
Repeated header6786.285.2+1.17th
References5586.285.2+1.08th
Busy page1883.580.1+3.43rd

Source

LlamaParse Extract Agentic by source
Kind of resume# of resumesF1MedianGapRank
Everyday resumes21286.185.1+1.07th
Expert-built (Canva / LaTeX)3283.783.6+0.18th
Resume-builder templates1083.281.2+2.08th
Scanned & rotated1991.682.9+8.72nd
Distractor layouts1888.385.3+3.03rd

Layout

LlamaParse Extract Agentic by layout
Kind of resume# of resumesF1MedianGapRank
Single column19585.284.0+1.25th
Two-column2389.186.8+2.34th
Two-column sidebar5484.080.0+4.12nd
Three-column2085.484.8+0.68th
Table grid895.294.1+1.25th

Length

LlamaParse Extract Agentic by length
Kind of resume# of resumesF1MedianGapRank
Short14889.487.1+2.22nd
Medium9084.484.7−0.210th
Long6278.074.6+3.44th

By section of the resume

F1 by resume section, against the median system.

LlamaParse Extract Agentic by resume section
SectionResumesF1MedianGap
Contact details30094.593.5+1.0
Experience30089.686.3+3.3
Education29791.389.4+1.9
Summary29084.581.7+2.8
Certifications25280.981.1−0.2
Awards9380.375.4+4.9
Skills28695.192.1+3.0
Languages16784.284.3−0.2
Custom sections20852.149.9+2.2
Interests6493.091.7+1.2
Volunteering7787.784.5+3.1
Profiles3567.666.7+0.9
Projects7886.283.2+3.0
Publications5180.578.6+2.0

On the published resumes

Best and worst of the 20 public resumes.

LlamaParse Extract Agentic on published resumes
ResumeWhat makes it hardF1
Head of Events & Public Space OperationsTable grid with emoji bullets98.4
Head of Developer Experience (Internal title: Jedi Master of Analytics) Unusual job titles91.9
Senior Controls EngineerFive custom sections89.6
Senior ocularist and practice ownerHandwritten resume, six pages70.4
Senior Full-Stack EngineerLaTeX CV with a running header66.9
Senior Game DesignerGame designer's Canva resume64.8

How it was run

LlamaParse Extract API, extraction mode agentic. Same PDF, schema and instructions as every other system, one run per resume (4 November 2026).

LlamaParse Extract (formerly LlamaExtract) is the structured-extraction API of LlamaIndex’s LlamaParse platform.

LlamaParse Extract Agentic on LlamaIndex’s site

Reproduce it with the open-source harness (GitHub):

$ pip install "resume-bench[runners] @ git+https://github.com/careerflow/ResumeBench"
$ hf auth login   # the dataset is gated: accept its terms on Hugging Face first
$ resume-bench download
$ resume-bench run llamaextract_agentic --split test
$ resume-bench grade llamaextract_agentic --split test

Closest alternatives

Nearest in F1.

Embed this result

Shows its current rank. Updates with each run.

LlamaParse Extract Agentic: #3 on ResumeBench resume extraction benchmark
<a href="https://www.resumebench.ai/systems/llamaparse-extract-agentic"><img src="https://www.resumebench.ai/badge/llamaparse-extract-agentic.svg" alt="LlamaParse Extract Agentic: #3 on ResumeBench resume extraction benchmark" height="28"></a>

Choosing a resume extraction model

All 18 models · Leaderboard

Frequently asked questions (FAQs)

How accurate is LlamaParse Extract Agentic at resume extraction?

85.5 F1 on ResumeBench's 300 resumes, 3rd of 18 models.

How much does LlamaParse Extract Agentic cost to parse resumes?

$23 per 1,000 pages, about $0.064 per resume.

How fast is LlamaParse Extract Agentic?

55 seconds per resume (median).

Which resumes is LlamaParse Extract Agentic best at?

Handwritten, image only and scanned & rotated. Weakest on medium.

What is a cheaper alternative to LlamaParse Extract Agentic?

Qwen 3.8 Flash: 84.1 F1 for $2.98 per 1,000 pages.