Qwen 3.5 9B vs Qwen 3.8 27B: which is better at resume extraction?
Qwen 3.8 27B leads Qwen 3.5 9B by 7.7 points (83.8 vs 76.0 F1). Cost per 1,000 pages: $1.31 vs $9.59. Median time per resume: 138 s vs 36 s.
Qwen 3.5 9B wins on 63 of 299 resumes, Qwen 3.8 27B on 218, 18 tied.
76.03
Qwen 3.5 9B
A · Vision LLM
83.86
Qwen 3.8 27B
B · Vision LLM
Qwen 3.8 27B is ahead by 7.83 points on these 299 resumes.
- Many entries 82A 73.9 · B 82.0−8.1
- Long bullet lists 100A 73.2 · B 81.0−7.8
- Split across pages 84A 73.4 · B 82.1−8.7
- Big skills list 59A 76.3 · B 83.3−7.0
- Custom sections 104A 73.0 · B 81.2−8.2
- Combined headings 37A 72.2 · B 83.7−11.5
- Promotions 88A 74.8 · B 83.2−8.4
- Partial dates 138A 75.6 · B 83.6−8.0
- Overlapping jobs 69A 72.2 · B 82.4−10.2
- Accented text 13A 83.7 · B 84.9−1.2
Side by side
| Qwen 3.5 9B | Qwen 3.8 27B | |
|---|---|---|
| Type | Vision LLM | Vision LLM |
| Maker | Alibaba | Alibaba |
| Rank | 15 of 18 | 9 of 18 |
| F1 | 76.0 | 83.8 |
| Precision | 78.9 | 85.4 |
| Recall | 80.3 | 85.5 |
| Cost per 1,000 pages | $1.31 | $9.59 |
| Median time per resume | 138 s | 36 s |
| Resumes won | 63 | 218 |
Where each one wins
Where the two differ most.
| Kind of resume | # of resumes | Qwen 3.5 9B | Qwen 3.8 27B | Gap |
|---|---|---|---|---|
| Scanned & rotated (source) | 19 | 64.5 | 78.4 | −13.9 |
| Busy page (noise) | 18 | 67.6 | 80.4 | −12.8 |
| Rotated & warped pages (distractors) | 21 | 66.5 | 78.7 | −12.3 |
| Hidden text & prompt injections (distractors) | 14 | 70.2 | 82.4 | −12.2 |
| Combined headings (extraction challenges) | 37 | 72.2 | 83.7 | −11.5 |
| Typos kept on purpose (distractors) | 37 | 72.2 | 82.9 | −10.7 |
| Overlapping jobs (extraction challenges) | 69 | 72.2 | 82.4 | −10.2 |
| Wording traps (distractors) | 16 | 77.7 | 87.6 | −9.9 |
| Three-column (layout) | 20 | 75.8 | 85.5 | −9.7 |
| Image only (reading the page) | 28 | 61.2 | 70.5 | −9.3 |
| Photo (reading the page) | 45 | 74.9 | 84.1 | −9.2 |
| Short (length) | 148 | 78.1 | 87.1 | −8.9 |
| Section | Resumes | Qwen 3.5 9B | Qwen 3.8 27B | Gap |
|---|---|---|---|---|
| Contact details | 299 | 92.6 | 93.2 | −0.6 |
| Experience | 299 | 77.2 | 86.1 | −8.9 |
| Education | 296 | 80.8 | 87.2 | −6.4 |
| Summary | 290 | 77.9 | 82.0 | −4.1 |
| Certifications | 246 | 74.7 | 82.7 | −8.0 |
| Awards | 127 | 55.2 | 78.2 | −23.0 |
| Skills | 293 | 83.2 | 92.8 | −9.5 |
| Languages | 172 | 80.0 | 85.4 | −5.4 |
| Custom sections | 214 | 36.4 | 51.5 | −15.1 |
| Interests | 68 | 84.3 | 94.6 | −10.3 |
| Volunteering | 80 | 80.2 | 84.7 | −4.5 |
| Profiles | 38 | 51.3 | 74.7 | −23.4 |
| Projects | 79 | 75.5 | 86.5 | −11.0 |
| Publications | 53 | 76.3 | 77.2 | −0.9 |
Published resumes they disagree on most
Open one to see both outputs field by field.
| Resume | Qwen 3.5 9B | Qwen 3.8 27B | Gap |
|---|---|---|---|
| Supervisory Human Resources Specialist, GS-0201-15: Resume-builder template | 54.9 | 88.5 | −33.6 |
| Head of Developer Experience (Internal title: Jedi Master of Analytics) : Unusual job titles | 77.2 | 96.3 | −19.1 |
| Senior Game Designer: Game designer's Canva resume | 55.1 | 74.0 | −18.9 |
| Postdoctoral Researcher: Academic CV, seven pages | 71.2 | 87.0 | −15.8 |
Which to pick
Qwen 3.8 27B is more accurate. Qwen 3.5 9B costs 86% less for 7.7 fewer points.
Qwen 3.8 27B does better on scanned & rotated, busy page and rotated & warped pages.
Qwen 3.5 9B against every other model
- Qwen 3.5 9B vs LlamaParse Extract Agentic Plus
- Qwen 3.5 9B vs LlamaParse Extract Agentic
- Qwen 3.5 9B vs LlamaParse Extract Cost-Effective
- Qwen 3.5 9B vs Reducto Deep Extract
- Qwen 3.5 9B vs Reducto Extract
- Qwen 3.5 9B vs Extend Max Context
- Qwen 3.5 9B vs Extend Extract
- Qwen 3.5 9B vs GPT-5.6
- Qwen 3.5 9B vs GPT-5.5
- Qwen 3.5 9B vs GPT-5.4 Nano
- Qwen 3.5 9B vs Claude Opus 4.8
- Qwen 3.5 9B vs Gemini 3.5 Flash
- Qwen 3.5 9B vs Gemma 4 26B
- Qwen 3.5 9B vs Qwen 3.8 Flash
- Qwen 3.5 9B vs Qwen 3.5 35B
- Qwen 3.5 9B vs GLM 5.3 Flash
Qwen 3.8 27B against every other model
- Qwen 3.8 27B vs LlamaParse Extract Agentic Plus
- Qwen 3.8 27B vs LlamaParse Extract Agentic
- Qwen 3.8 27B vs LlamaParse Extract Cost-Effective
- Qwen 3.8 27B vs Reducto Deep Extract
- Qwen 3.8 27B vs Reducto Extract
- Qwen 3.8 27B vs Extend Max Context
- Qwen 3.8 27B vs Extend Extract
- Qwen 3.8 27B vs GPT-5.6
- Qwen 3.8 27B vs GPT-5.5
- Qwen 3.8 27B vs GPT-5.4 Nano
- Qwen 3.8 27B vs Claude Opus 4.8
- Qwen 3.8 27B vs Gemini 3.5 Flash
- Qwen 3.8 27B vs Gemma 4 26B
- Qwen 3.8 27B vs Qwen 3.8 Flash
- Qwen 3.8 27B vs Qwen 3.5 35B
- Qwen 3.8 27B vs GLM 5.3 Flash
Popular comparisons
- LlamaParse Extract Agentic Plus vs Reducto Deep Extract
- Extend Max Context vs Reducto Deep Extract
- Extend Max Context vs LlamaParse Extract Agentic Plus
- Claude Opus 4.8 vs GPT-5.6
- Gemini 3.5 Flash vs GPT-5.6
- Claude Opus 4.8 vs Gemini 3.5 Flash
- LlamaParse Extract Agentic vs LlamaParse Extract Agentic Plus
- Reducto Deep Extract vs Reducto Extract
- Gemma 4 26B vs Qwen 3.8 27B
- Document APIs vs vision LLMs