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
| Kind of resume | # of resumes | F1 | Median | Gap | Rank |
|---|---|---|---|---|---|
| Handwritten | 9 | 63.5 | 51.5 | +12.0 | 2nd |
| Image only | 28 | 81.5 | 71.4 | +10.1 | 2nd |
| Sidebar or grid | 82 | 85.5 | 81.9 | +3.5 | 2nd |
| Photo | 45 | 84.7 | 84.0 | +0.6 | 8th |
Distractors
| Kind of resume | # of resumes | F1 | Median | Gap | Rank |
|---|---|---|---|---|---|
| Rotated & warped pages | 21 | 90.9 | 82.9 | +8.0 | 2nd |
| Hidden text & prompt injections | 14 | 91.6 | 84.5 | +7.0 | 2nd |
| Content as images | 8 | 91.6 | 88.2 | +3.4 | 4th |
| Wording traps | 16 | 87.7 | 84.5 | +3.2 | 3rd |
| Typos kept on purpose | 37 | 90.0 | 83.7 | +6.4 | 2nd |
Extraction challenges
| Kind of resume | # of resumes | F1 | Median | Gap | Rank |
|---|---|---|---|---|---|
| Many entries | 83 | 83.0 | 81.9 | +1.2 | 8th |
| Long bullet lists | 101 | 83.1 | 81.4 | +1.7 | 8th |
| Split across pages | 85 | 83.8 | 82.2 | +1.6 | 7th |
| Big skills list | 60 | 83.7 | 83.6 | +0.1 | 9th |
| Custom sections | 105 | 82.5 | 80.8 | +1.7 | 5th |
| Combined headings | 37 | 83.2 | 82.7 | +0.5 | 7th |
| Promotions | 88 | 85.5 | 83.5 | +2.0 | 5th |
| Partial dates | 138 | 85.6 | 83.4 | +2.1 | 3rd |
| Overlapping jobs | 69 | 83.5 | 81.7 | +1.8 | 5th |
| Accented text | 13 | 91.1 | 85.5 | +5.6 | 2nd |
Noise
| Kind of resume | # of resumes | F1 | Median | Gap | Rank |
|---|---|---|---|---|---|
| Repeated header | 67 | 86.2 | 85.2 | +1.1 | 7th |
| References | 55 | 86.2 | 85.2 | +1.0 | 8th |
| Busy page | 18 | 83.5 | 80.1 | +3.4 | 3rd |
Source
| Kind of resume | # of resumes | F1 | Median | Gap | Rank |
|---|---|---|---|---|---|
| Everyday resumes | 212 | 86.1 | 85.1 | +1.0 | 7th |
| Expert-built (Canva / LaTeX) | 32 | 83.7 | 83.6 | +0.1 | 8th |
| Resume-builder templates | 10 | 83.2 | 81.2 | +2.0 | 8th |
| Scanned & rotated | 19 | 91.6 | 82.9 | +8.7 | 2nd |
| Distractor layouts | 18 | 88.3 | 85.3 | +3.0 | 3rd |
Layout
| Kind of resume | # of resumes | F1 | Median | Gap | Rank |
|---|---|---|---|---|---|
| Single column | 195 | 85.2 | 84.0 | +1.2 | 5th |
| Two-column | 23 | 89.1 | 86.8 | +2.3 | 4th |
| Two-column sidebar | 54 | 84.0 | 80.0 | +4.1 | 2nd |
| Three-column | 20 | 85.4 | 84.8 | +0.6 | 8th |
| Table grid | 8 | 95.2 | 94.1 | +1.2 | 5th |
Length
| Kind of resume | # of resumes | F1 | Median | Gap | Rank |
|---|---|---|---|---|---|
| Short | 148 | 89.4 | 87.1 | +2.2 | 2nd |
| Medium | 90 | 84.4 | 84.7 | −0.2 | 10th |
| Long | 62 | 78.0 | 74.6 | +3.4 | 4th |
By section of the resume
F1 by resume section, against the median system.
| Section | Resumes | F1 | Median | Gap |
|---|---|---|---|---|
| Contact details | 300 | 94.5 | 93.5 | +1.0 |
| Experience | 300 | 89.6 | 86.3 | +3.3 |
| Education | 297 | 91.3 | 89.4 | +1.9 |
| Summary | 290 | 84.5 | 81.7 | +2.8 |
| Certifications | 252 | 80.9 | 81.1 | −0.2 |
| Awards | 93 | 80.3 | 75.4 | +4.9 |
| Skills | 286 | 95.1 | 92.1 | +3.0 |
| Languages | 167 | 84.2 | 84.3 | −0.2 |
| Custom sections | 208 | 52.1 | 49.9 | +2.2 |
| Interests | 64 | 93.0 | 91.7 | +1.2 |
| Volunteering | 77 | 87.7 | 84.5 | +3.1 |
| Profiles | 35 | 67.6 | 66.7 | +0.9 |
| Projects | 78 | 86.2 | 83.2 | +3.0 |
| Publications | 51 | 80.5 | 78.6 | +2.0 |
On the published resumes
Best and worst of the 20 public resumes.
| Resume | What makes it hard | F1 |
|---|---|---|
| Head of Events & Public Space Operations | Table grid with emoji bullets | 98.4 |
| Head of Developer Experience (Internal title: Jedi Master of Analytics) | Unusual job titles | 91.9 |
| Senior Controls Engineer | Five custom sections | 89.6 |
| Senior ocularist and practice owner | Handwritten resume, six pages | 70.4 |
| Senior Full-Stack Engineer | LaTeX CV with a running header | 66.9 |
| Senior Game Designer | Game designer's Canva resume | 64.8 |
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.
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 testClosest alternatives
Nearest in F1.
| Model | F1 | Per 1,000 pages | Median time | Head to head |
|---|---|---|---|---|
| Reducto Deep Extract | 85.6 | $40 | 65 s | LlamaParse Extract Agentic vs Reducto Deep Extract |
| Claude Opus 4.8 | 85.5 | $90 | 31 s | LlamaParse Extract Agentic vs Claude Opus 4.8 |
| Gemini 3.5 Flash | 85.5 | $35 | 43 s | LlamaParse Extract Agentic vs Gemini 3.5 Flash |
| GPT-5.6 | 84.7 | $74 | 38 s | LlamaParse Extract Agentic vs GPT-5.6 |
LlamaParse Extract Agentic against every other model
- LlamaParse Extract Agentic vs LlamaParse Extract Agentic Plus
- LlamaParse Extract Agentic vs LlamaParse Extract Cost-Effective
- LlamaParse Extract Agentic vs Reducto Extract
- LlamaParse Extract Agentic vs Extend Max Context
- LlamaParse Extract Agentic vs Extend Extract
- LlamaParse Extract Agentic vs GPT-5.5
- LlamaParse Extract Agentic vs GPT-5.4 Nano
- LlamaParse Extract Agentic vs Gemma 4 26B
- LlamaParse Extract Agentic vs Qwen 3.8 Flash
- LlamaParse Extract Agentic vs Qwen 3.8 27B
- LlamaParse Extract Agentic vs Qwen 3.5 35B
- LlamaParse Extract Agentic vs Qwen 3.5 9B
- LlamaParse Extract Agentic vs GLM 5.3 Flash
Embed this result
Shows its current rank. Updates with each run.