Gemma 4 26B vs Qwen 3.8 27B: which is better at resume extraction?
Qwen 3.8 27B leads Gemma 4 26B by 20.9 points (83.8 vs 62.8 F1). Cost per 1,000 pages: $0.56 vs $9.59. Median time per resume: 29 s vs 36 s.
Gemma 4 26B wins on 19 of 300 resumes, Qwen 3.8 27B on 277, 4 tied.
62.82
Gemma 4 26B
A · Vision LLM
83.77
Qwen 3.8 27B
B · Vision LLM
Qwen 3.8 27B is ahead by 20.95 points on these 300 resumes.
- Many entries 83A 58.0 · B 81.6−23.6
- Long bullet lists 101A 57.4 · B 80.8−23.4
- Split across pages 85A 60.0 · B 81.8−21.8
- Big skills list 60A 61.9 · B 82.9−20.9
- Custom sections 105A 58.9 · B 81.0−22.1
- Combined headings 37A 60.9 · B 83.7−22.8
- Promotions 88A 59.5 · B 83.2−23.7
- Partial dates 138A 62.9 · B 83.6−20.7
- Overlapping jobs 69A 57.5 · B 82.4−24.9
- Accented text 13A 59.3 · B 84.9−25.7
Side by side
| Gemma 4 26B | Qwen 3.8 27B | |
|---|---|---|
| Type | Vision LLM | Vision LLM |
| Maker | Alibaba | |
| Rank | 16 of 18 | 9 of 18 |
| F1 | 62.8 | 83.8 |
| Precision | 69.1 | 85.4 |
| Recall | 66.2 | 85.5 |
| Cost per 1,000 pages | $0.56 | $9.59 |
| Median time per resume | 29 s | 36 s |
| Resumes won | 19 | 277 |
Where each one wins
Where the two differ most.
| Kind of resume | # of resumes | Gemma 4 26B | Qwen 3.8 27B | Gap |
|---|---|---|---|---|
| Scanned & rotated (source) | 19 | 40.7 | 78.4 | −37.7 |
| Rotated & warped pages (distractors) | 21 | 43.4 | 78.7 | −35.3 |
| Hidden text & prompt injections (distractors) | 14 | 48.6 | 82.4 | −33.8 |
| Typos kept on purpose (distractors) | 37 | 51.3 | 82.9 | −31.7 |
| Resume-builder templates (source) | 10 | 58.6 | 86.7 | −28.1 |
| Wording traps (distractors) | 16 | 60.6 | 87.6 | −27.0 |
| Accented text (extraction challenges) | 13 | 59.3 | 84.9 | −25.7 |
| Image only (reading the page) | 28 | 45.1 | 70.5 | −25.4 |
| Distractor layouts (source) | 18 | 62.4 | 87.6 | −25.3 |
| Overlapping jobs (extraction challenges) | 69 | 57.5 | 82.4 | −24.9 |
| Content as images (distractors) | 8 | 65.6 | 89.5 | −23.9 |
| Promotions (extraction challenges) | 88 | 59.5 | 83.2 | −23.7 |
| Section | Resumes | Gemma 4 26B | Qwen 3.8 27B | Gap |
|---|---|---|---|---|
| Contact details | 300 | 84.6 | 93.2 | −8.6 |
| Experience | 300 | 56.0 | 86.2 | −30.2 |
| Education | 297 | 68.0 | 87.2 | −19.2 |
| Summary | 290 | 35.5 | 81.7 | −46.2 |
| Certifications | 244 | 63.0 | 82.7 | −19.8 |
| Awards | 110 | 55.7 | 78.4 | −22.7 |
| Skills | 290 | 81.9 | 92.7 | −10.8 |
| Languages | 171 | 80.9 | 85.4 | −4.4 |
| Custom sections | 223 | 32.3 | 51.2 | −18.9 |
| Interests | 67 | 71.8 | 94.6 | −22.7 |
| Volunteering | 78 | 59.4 | 84.7 | −25.3 |
| Profiles | 37 | 57.2 | 72.5 | −15.3 |
| Projects | 83 | 55.0 | 85.5 | −30.5 |
| Publications | 53 | 54.5 | 77.2 | −22.7 |
Published resumes they disagree on most
Open one to see both outputs field by field.
| Resume | Gemma 4 26B | Qwen 3.8 27B | Gap |
|---|---|---|---|
| Supervisory Human Resources Specialist, GS-0201-15: Resume-builder template | 29.4 | 88.5 | −59.1 |
| Head of Developer Experience (Internal title: Jedi Master of Analytics) : Unusual job titles | 53.7 | 96.3 | −42.6 |
| Creative Director: Accented text in a styled CV | 39.3 | 78.3 | −39.1 |
| Deputy Head of Mission: Rotated scan with hidden identity text | 34.1 | 65.6 | −31.5 |
Which to pick
Qwen 3.8 27B is more accurate. Gemma 4 26B costs 94% less for 20.9 fewer points.
Qwen 3.8 27B does better on scanned & rotated, rotated & warped pages and hidden text & prompt injections.
Gemma 4 26B against every other model
- Gemma 4 26B vs LlamaParse Extract Agentic Plus
- Gemma 4 26B vs LlamaParse Extract Agentic
- Gemma 4 26B vs LlamaParse Extract Cost-Effective
- Gemma 4 26B vs Reducto Deep Extract
- Gemma 4 26B vs Reducto Extract
- Gemma 4 26B vs Extend Max Context
- Gemma 4 26B vs Extend Extract
- Gemma 4 26B vs GPT-5.6
- Gemma 4 26B vs GPT-5.5
- Gemma 4 26B vs GPT-5.4 Nano
- Gemma 4 26B vs Claude Opus 4.8
- Gemma 4 26B vs Gemini 3.5 Flash
- Gemma 4 26B vs Qwen 3.8 Flash
- Gemma 4 26B vs Qwen 3.5 35B
- Gemma 4 26B vs Qwen 3.5 9B
- Gemma 4 26B 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 Qwen 3.8 Flash
- Qwen 3.8 27B vs Qwen 3.5 35B
- Qwen 3.8 27B vs Qwen 3.5 9B
- 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
- Document APIs vs vision LLMs