Extend · Document API
Extend Extract on ResumeBench: 81.0 F1, rank 13 of 18
Extend Extract ranks 13th of 18 on ResumeBench: 81.0 F1, $12 per 1,000 pages, 31 seconds per resume. It is 6.2 points behind LlamaParse Extract Agentic Plus. Qwen 3.8 Flash scores 84.1 for $2.98.
7th of 7 document APIs. Strongest on handwritten (+1.0 points), content as images (+0.0 points) and distractor layouts (−0.1 points). Weakest on expert-built (canva / latex) (−6.5 points), two-column (−5.5 points) and photo (−5.5 points).
- Rank
- 13/18
- F1
- 81.0
- Precision
- 82.9
- Recall
- 83.2
- Perfect resumes
- 16/300
- Per 1,000 pages
- $12
- Median per resume
- 31 s
Averaged over 300 resumes. Dates (scored separately): 96.6. 90% finish within 50 s. Cost from list price.
Where Extend Extract 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 | 52.5 | 51.5 | +1.0 | 8th |
| Image only | 28 | 70.7 | 71.4 | −0.6 | 10th |
| Sidebar or grid | 82 | 80.0 | 81.9 | −2.0 | 13th |
| Photo | 45 | 78.5 | 84.0 | −5.5 | 14th |
Distractors
| Kind of resume | # of resumes | F1 | Median | Gap | Rank |
|---|---|---|---|---|---|
| Rotated & warped pages | 21 | 79.8 | 82.9 | −3.1 | 10th |
| Hidden text & prompt injections | 14 | 82.4 | 84.5 | −2.1 | 10th |
| Content as images | 8 | 88.2 | 88.2 | +0.0 | 9th |
| Wording traps | 16 | 82.5 | 84.5 | −2.0 | 11th |
| Typos kept on purpose | 37 | 82.3 | 83.7 | −1.4 | 11th |
Extraction challenges
| Kind of resume | # of resumes | F1 | Median | Gap | Rank |
|---|---|---|---|---|---|
| Many entries | 83 | 80.2 | 81.9 | −1.6 | 11th |
| Long bullet lists | 101 | 79.8 | 81.4 | −1.7 | 13th |
| Split across pages | 85 | 79.6 | 82.2 | −2.6 | 13th |
| Big skills list | 60 | 81.4 | 83.6 | −2.2 | 13th |
| Custom sections | 105 | 77.3 | 80.8 | −3.5 | 13th |
| Combined headings | 37 | 78.0 | 82.7 | −4.7 | 13th |
| Promotions | 88 | 81.7 | 83.5 | −1.8 | 13th |
| Partial dates | 138 | 80.1 | 83.4 | −3.4 | 12th |
| Overlapping jobs | 69 | 81.2 | 81.7 | −0.5 | 11th |
| Accented text | 13 | 84.3 | 85.5 | −1.3 | 12th |
Noise
| Kind of resume | # of resumes | F1 | Median | Gap | Rank |
|---|---|---|---|---|---|
| Repeated header | 67 | 82.2 | 85.2 | −3.0 | 13th |
| References | 55 | 84.8 | 85.2 | −0.4 | 11th |
| Busy page | 18 | 78.2 | 80.1 | −1.9 | 11th |
Source
| Kind of resume | # of resumes | F1 | Median | Gap | Rank |
|---|---|---|---|---|---|
| Everyday resumes | 212 | 82.6 | 85.1 | −2.6 | 13th |
| Expert-built (Canva / LaTeX) | 32 | 77.1 | 83.6 | −6.5 | 15th |
| Resume-builder templates | 10 | 80.1 | 81.2 | −1.0 | 12th |
| Scanned & rotated | 19 | 79.5 | 82.9 | −3.4 | 10th |
| Distractor layouts | 18 | 85.2 | 85.3 | −0.1 | 10th |
Layout
| Kind of resume | # of resumes | F1 | Median | Gap | Rank |
|---|---|---|---|---|---|
| Single column | 195 | 81.4 | 84.0 | −2.6 | 12th |
| Two-column | 23 | 81.3 | 86.8 | −5.5 | 13th |
| Two-column sidebar | 54 | 78.3 | 80.0 | −1.7 | 12th |
| Three-column | 20 | 80.4 | 84.8 | −4.4 | 12th |
| Table grid | 8 | 89.8 | 94.1 | −4.2 | 15th |
Length
| Kind of resume | # of resumes | F1 | Median | Gap | Rank |
|---|---|---|---|---|---|
| Short | 148 | 83.7 | 87.1 | −3.4 | 13th |
| Medium | 90 | 81.6 | 84.7 | −3.1 | 13th |
| Long | 62 | 73.6 | 74.6 | −1.0 | 11th |
By section of the resume
F1 by resume section, against the median system.
| Section | Resumes | F1 | Median | Gap |
|---|---|---|---|---|
| Contact details | 300 | 91.1 | 93.5 | −2.5 |
| Experience | 300 | 81.0 | 86.3 | −5.4 |
| Education | 297 | 90.2 | 89.4 | +0.8 |
| Summary | 290 | 80.7 | 81.7 | −1.0 |
| Certifications | 254 | 75.7 | 81.1 | −5.4 |
| Awards | 99 | 73.8 | 75.4 | −1.5 |
| Skills | 287 | 88.7 | 92.1 | −3.4 |
| Languages | 167 | 85.1 | 84.3 | +0.8 |
| Custom sections | 212 | 44.4 | 49.9 | −5.5 |
| Interests | 64 | 90.9 | 91.7 | −0.8 |
| Volunteering | 77 | 84.5 | 84.5 | −0.1 |
| Profiles | 36 | 42.6 | 66.7 | −24.1 |
| Projects | 82 | 80.2 | 83.2 | −3.0 |
| Publications | 51 | 85.1 | 78.6 | +6.5 |
On the published resumes
Best and worst of the 20 public resumes.
| Resume | What makes it hard | F1 |
|---|---|---|
| Senior Manufacturing Engineer | Third-person prose, no bullets | 94.5 |
| Lead Respiratory Therapist (Adult ICU) | Long resume, 34 entries | 88.3 |
| Head of Events & Public Space Operations | Table grid with emoji bullets | 88.3 |
| Head of Developer Experience (Internal title: Jedi Master of Analytics) | Unusual job titles | 55.1 |
| Postdoctoral Researcher | Academic CV, seven pages | 54.2 |
| Senior ocularist and practice owner | Handwritten resume, six pages | 44.0 |
How it was run
Extend extraction, default settings. Same PDF, schema and instructions as every other system, one run per resume (4 November 2026).
Run the open-source harness: list its pipelines, or grade your own predictions (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 providers
$ resume-bench grade-file my_predictions.jsonl --split testClosest alternatives
Nearest in F1.
| Model | F1 | Per 1,000 pages | Median time | Head to head |
|---|---|---|---|---|
| Qwen 3.8 27B | 83.8 | $9.59 | 36 s | Extend Extract vs Qwen 3.8 27B |
| LlamaParse Extract Cost-Effective | 83.5 | $8.18 | 29 s | Extend Extract vs LlamaParse Extract Cost-Effective |
| GPT-5.5 | 83.5 | $87 | 48 s | Extend Extract vs GPT-5.5 |
| Reducto Extract | 81.7 | $20 | 16 s | Extend Extract vs Reducto Extract |
Extend Extract against every other model
- Extend Extract vs LlamaParse Extract Agentic Plus
- Extend Extract vs LlamaParse Extract Agentic
- Extend Extract vs Reducto Deep Extract
- Extend Extract vs Extend Max Context
- Extend Extract vs GPT-5.6
- Extend Extract vs GPT-5.4 Nano
- Extend Extract vs Claude Opus 4.8
- Extend Extract vs Gemini 3.5 Flash
- Extend Extract vs Gemma 4 26B
- Extend Extract vs Qwen 3.8 Flash
- Extend Extract vs Qwen 3.5 35B
- Extend Extract vs Qwen 3.5 9B
- Extend Extract vs GLM 5.3 Flash
Embed this result
Shows its current rank. Updates with each run.