LlamaIndex · Document API
LlamaParse Extract Cost-Effective on ResumeBench: 83.5 F1, rank 10 of 18
LlamaParse Extract Cost-Effective ranks 10th of 18 on ResumeBench: 83.5 F1, $8.18 per 1,000 pages, 29 seconds per resume. It is 3.6 points behind LlamaParse Extract Agentic Plus. Qwen 3.8 Flash scores 84.1 for $2.98.
5th of 7 document APIs. Strongest on scanned & rotated (+7.4 points), rotated & warped pages (+6.7 points) and hidden text & prompt injections (+6.0 points). Weakest on busy page (−3.2 points), references (−2.3 points) and many entries (−2.2 points).
- Rank
- 10/18
- F1
- 83.5
- Precision
- 85.1
- Recall
- 85.3
- Perfect resumes
- 26/300
- Per 1,000 pages
- $8.18
- Median per resume
- 29 s
Averaged over 300 resumes. Dates (scored separately): 97.9. 90% finish within 42 s.
Where LlamaParse Extract Cost-Effective 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 | 49.6 | 51.5 | −1.9 | 15th |
| Image only | 28 | 76.3 | 71.4 | +4.9 | 6th |
| Sidebar or grid | 82 | 84.3 | 81.9 | +2.3 | 6th |
| Photo | 45 | 83.1 | 84.0 | −0.9 | 11th |
Distractors
| Kind of resume | # of resumes | F1 | Median | Gap | Rank |
|---|---|---|---|---|---|
| Rotated & warped pages | 21 | 89.6 | 82.9 | +6.7 | 4th |
| Hidden text & prompt injections | 14 | 90.5 | 84.5 | +6.0 | 4th |
| Content as images | 8 | 88.2 | 88.2 | −0.0 | 10th |
| Wording traps | 16 | 85.0 | 84.5 | +0.5 | 8th |
| Typos kept on purpose | 37 | 88.0 | 83.7 | +4.3 | 4th |
Extraction challenges
| Kind of resume | # of resumes | F1 | Median | Gap | Rank |
|---|---|---|---|---|---|
| Many entries | 83 | 79.6 | 81.9 | −2.2 | 12th |
| Long bullet lists | 101 | 80.4 | 81.4 | −1.1 | 11th |
| Split across pages | 85 | 81.8 | 82.2 | −0.4 | 11th |
| Big skills list | 60 | 84.1 | 83.6 | +0.5 | 8th |
| Custom sections | 105 | 80.4 | 80.8 | −0.4 | 11th |
| Combined headings | 37 | 82.8 | 82.7 | +0.1 | 8th |
| Promotions | 88 | 82.9 | 83.5 | −0.6 | 11th |
| Partial dates | 138 | 82.5 | 83.4 | −1.0 | 11th |
| Overlapping jobs | 69 | 79.7 | 81.7 | −2.0 | 13th |
| Accented text | 13 | 87.0 | 85.5 | +1.5 | 5th |
Noise
| Kind of resume | # of resumes | F1 | Median | Gap | Rank |
|---|---|---|---|---|---|
| Repeated header | 67 | 84.3 | 85.2 | −0.8 | 11th |
| References | 55 | 82.8 | 85.2 | −2.3 | 13th |
| Busy page | 18 | 76.9 | 80.1 | −3.2 | 12th |
Source
| Kind of resume | # of resumes | F1 | Median | Gap | Rank |
|---|---|---|---|---|---|
| Everyday resumes | 212 | 84.6 | 85.1 | −0.6 | 11th |
| Expert-built (Canva / LaTeX) | 32 | 81.8 | 83.6 | −1.8 | 12th |
| Resume-builder templates | 10 | 80.3 | 81.2 | −0.9 | 11th |
| Scanned & rotated | 19 | 90.4 | 82.9 | +7.4 | 4th |
| Distractor layouts | 18 | 85.5 | 85.3 | +0.1 | 9th |
Layout
| Kind of resume | # of resumes | F1 | Median | Gap | Rank |
|---|---|---|---|---|---|
| Single column | 195 | 82.8 | 84.0 | −1.1 | 11th |
| Two-column | 23 | 86.7 | 86.8 | −0.1 | 10th |
| Two-column sidebar | 54 | 82.3 | 80.0 | +2.3 | 6th |
| Three-column | 20 | 85.9 | 84.8 | +1.1 | 4th |
| Table grid | 8 | 93.7 | 94.1 | −0.4 | 10th |
Length
| Kind of resume | # of resumes | F1 | Median | Gap | Rank |
|---|---|---|---|---|---|
| Short | 148 | 87.7 | 87.1 | +0.6 | 7th |
| Medium | 90 | 83.0 | 84.7 | −1.7 | 11th |
| Long | 62 | 74.3 | 74.6 | −0.3 | 10th |
By section of the resume
F1 by resume section, against the median system.
| Section | Resumes | F1 | Median | Gap |
|---|---|---|---|---|
| Contact details | 300 | 93.6 | 93.5 | +0.1 |
| Experience | 300 | 84.5 | 86.3 | −1.8 |
| Education | 297 | 87.6 | 89.4 | −1.8 |
| Summary | 290 | 81.7 | 81.7 | +0.0 |
| Certifications | 246 | 82.5 | 81.1 | +1.4 |
| Awards | 93 | 84.1 | 75.4 | +8.8 |
| Skills | 286 | 94.3 | 92.1 | +2.2 |
| Languages | 167 | 82.2 | 84.3 | −2.1 |
| Custom sections | 215 | 48.8 | 49.9 | −1.1 |
| Interests | 64 | 92.1 | 91.7 | +0.4 |
| Volunteering | 77 | 87.4 | 84.5 | +2.8 |
| Profiles | 37 | 69.4 | 66.7 | +2.7 |
| Projects | 78 | 82.7 | 83.2 | −0.5 |
| Publications | 51 | 72.2 | 78.6 | −6.3 |
On the published resumes
Best and worst of the 20 public resumes.
| Resume | What makes it hard | F1 |
|---|---|---|
| Head of Developer Experience (Internal title: Jedi Master of Analytics) | Unusual job titles | 92.7 |
| Deputy Head of Mission | Rotated scan with hidden identity text | 88.9 |
| Head of Digital Marketing | Canva design with a photo | 85.6 |
| Senior Game Designer | Game designer's Canva resume | 55.0 |
| Senior ocularist and practice owner | Handwritten resume, six pages | 51.4 |
| senior Consultant - package Implementation | Handwritten resume | 44.1 |
How it was run
LlamaParse Extract API, extraction mode cost_effective. 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_cost_effective --split test
$ resume-bench grade llamaextract_cost_effective --split testClosest alternatives
Nearest in F1.
| Model | F1 | Per 1,000 pages | Median time | Head to head |
|---|---|---|---|---|
| Extend Max Context | 84.7 | $50 | 43 s | LlamaParse Extract Cost-Effective vs Extend Max Context |
| Qwen 3.8 Flash | 84.1 | $2.98 | 122 s | LlamaParse Extract Cost-Effective vs Qwen 3.8 Flash |
| Qwen 3.8 27B | 83.8 | $9.59 | 36 s | LlamaParse Extract Cost-Effective vs Qwen 3.8 27B |
| GPT-5.5 | 83.5 | $87 | 48 s | LlamaParse Extract Cost-Effective vs GPT-5.5 |
LlamaParse Extract Cost-Effective against every other model
- LlamaParse Extract Cost-Effective vs LlamaParse Extract Agentic Plus
- LlamaParse Extract Cost-Effective vs LlamaParse Extract Agentic
- LlamaParse Extract Cost-Effective vs Reducto Deep Extract
- LlamaParse Extract Cost-Effective vs Reducto Extract
- LlamaParse Extract Cost-Effective vs Extend Extract
- LlamaParse Extract Cost-Effective vs GPT-5.6
- LlamaParse Extract Cost-Effective vs GPT-5.4 Nano
- LlamaParse Extract Cost-Effective vs Claude Opus 4.8
- LlamaParse Extract Cost-Effective vs Gemini 3.5 Flash
- LlamaParse Extract Cost-Effective vs Gemma 4 26B
- LlamaParse Extract Cost-Effective vs Qwen 3.5 35B
- LlamaParse Extract Cost-Effective vs Qwen 3.5 9B
- LlamaParse Extract Cost-Effective vs GLM 5.3 Flash
Embed this result
Shows its current rank. Updates with each run.