OpenAI · Vision LLM
GPT-5.6 on ResumeBench: 84.7 F1, rank 6 of 18
GPT-5.6 ranks 6th of 18 on ResumeBench: 84.7 F1, $74 per 1,000 pages, 38 seconds per resume. It is 2.4 points behind LlamaParse Extract Agentic Plus. LlamaParse Extract Agentic Plus scores 87.2 for $54.
3rd of 11 vision LLMs. Strongest on content as images (+4.8 points), distractor layouts (+4.7 points) and wording traps (+4.6 points). Weakest on scanned & rotated (−11.4 points), rotated & warped pages (−9.7 points) and hidden text & prompt injections (−7.0 points).
- Rank
- 6/18
- F1
- 84.7
- Precision
- 85.9
- Recall
- 87.2
- Perfect resumes
- 36/300
- Per 1,000 pages
- $74
- Median per resume
- 38 s
Averaged over 300 resumes. Dates (scored separately): 97.6. 90% finish within 66 s.
Where GPT-5.6 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 | 53.2 | 51.5 | +1.8 | 6th |
| Image only | 28 | 65.4 | 71.4 | −5.9 | 13th |
| Sidebar or grid | 82 | 81.7 | 81.9 | −0.3 | 11th |
| Photo | 45 | 85.5 | 84.0 | +1.5 | 6th |
Distractors
| Kind of resume | # of resumes | F1 | Median | Gap | Rank |
|---|---|---|---|---|---|
| Rotated & warped pages | 21 | 73.2 | 82.9 | −9.7 | 13th |
| Hidden text & prompt injections | 14 | 77.5 | 84.5 | −7.0 | 13th |
| Content as images | 8 | 93.0 | 88.2 | +4.8 | 2nd |
| Wording traps | 16 | 89.1 | 84.5 | +4.6 | 1st |
| Typos kept on purpose | 37 | 80.6 | 83.7 | −3.1 | 12th |
Extraction challenges
| Kind of resume | # of resumes | F1 | Median | Gap | Rank |
|---|---|---|---|---|---|
| Many entries | 83 | 84.5 | 81.9 | +2.6 | 3rd |
| Long bullet lists | 101 | 84.9 | 81.4 | +3.5 | 2nd |
| Split across pages | 85 | 84.9 | 82.2 | +2.6 | 3rd |
| Big skills list | 60 | 85.2 | 83.6 | +1.5 | 6th |
| Custom sections | 105 | 82.2 | 80.8 | +1.3 | 6th |
| Combined headings | 37 | 84.1 | 82.7 | +1.4 | 4th |
| Promotions | 88 | 85.0 | 83.5 | +1.5 | 7th |
| Partial dates | 138 | 85.7 | 83.4 | +2.2 | 2nd |
| Overlapping jobs | 69 | 83.2 | 81.7 | +1.4 | 7th |
| Accented text | 13 | 86.5 | 85.5 | +1.0 | 7th |
Noise
| Kind of resume | # of resumes | F1 | Median | Gap | Rank |
|---|---|---|---|---|---|
| Repeated header | 67 | 86.8 | 85.2 | +1.7 | 4th |
| References | 55 | 87.9 | 85.2 | +2.7 | 5th |
| Busy page | 18 | 84.0 | 80.1 | +3.9 | 1st |
Source
| Kind of resume | # of resumes | F1 | Median | Gap | Rank |
|---|---|---|---|---|---|
| Everyday resumes | 212 | 86.8 | 85.1 | +1.7 | 4th |
| Expert-built (Canva / LaTeX) | 32 | 85.2 | 83.6 | +1.6 | 3rd |
| Resume-builder templates | 10 | 83.7 | 81.2 | +2.5 | 7th |
| Scanned & rotated | 19 | 71.6 | 82.9 | −11.4 | 13th |
| Distractor layouts | 18 | 90.1 | 85.3 | +4.7 | 1st |
Layout
| Kind of resume | # of resumes | F1 | Median | Gap | Rank |
|---|---|---|---|---|---|
| Single column | 195 | 85.5 | 84.0 | +1.6 | 3rd |
| Two-column | 23 | 89.0 | 86.8 | +2.1 | 5th |
| Two-column sidebar | 54 | 78.3 | 80.0 | −1.6 | 11th |
| Three-column | 20 | 85.6 | 84.8 | +0.7 | 6th |
| Table grid | 8 | 94.5 | 94.1 | +0.5 | 8th |
Length
| Kind of resume | # of resumes | F1 | Median | Gap | Rank |
|---|---|---|---|---|---|
| Short | 148 | 87.2 | 87.1 | +0.0 | 9th |
| Medium | 90 | 85.6 | 84.7 | +1.0 | 4th |
| Long | 62 | 77.7 | 74.6 | +3.0 | 5th |
By section of the resume
F1 by resume section, against the median system.
| Section | Resumes | F1 | Median | Gap |
|---|---|---|---|---|
| Contact details | 300 | 94.1 | 93.5 | +0.5 |
| Experience | 300 | 88.4 | 86.3 | +2.1 |
| Education | 297 | 91.9 | 89.4 | +2.5 |
| Summary | 290 | 83.4 | 81.7 | +1.7 |
| Certifications | 255 | 77.9 | 81.1 | −3.2 |
| Awards | 95 | 81.2 | 75.4 | +5.8 |
| Skills | 287 | 93.2 | 92.1 | +1.1 |
| Languages | 168 | 85.9 | 84.3 | +1.5 |
| Custom sections | 212 | 51.2 | 49.9 | +1.3 |
| Interests | 64 | 91.4 | 91.7 | −0.4 |
| Volunteering | 77 | 85.7 | 84.5 | +1.1 |
| Profiles | 37 | 69.4 | 66.7 | +2.7 |
| Projects | 80 | 88.2 | 83.2 | +5.0 |
| Publications | 51 | 77.8 | 78.6 | −0.8 |
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 | 93.6 |
| Creative Director | Accented text in a styled CV | 90.8 |
| Head of Events & Public Space Operations | Table grid with emoji bullets | 90.0 |
| Senior Full-Stack Engineer | LaTeX CV with a running header | 69.7 |
| senior Consultant - package Implementation | Handwritten resume | 47.2 |
| Senior ocularist and practice owner | Handwritten resume, six pages | 40.6 |
How it was run
OpenAI API, model gpt-5.6-sol, page images in, JSON schema out. 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 |
|---|---|---|---|---|
| Claude Opus 4.8 | 85.5 | $90 | 31 s | GPT-5.6 vs Claude Opus 4.8 |
| Gemini 3.5 Flash | 85.5 | $35 | 43 s | GPT-5.6 vs Gemini 3.5 Flash |
| Extend Max Context | 84.7 | $50 | 43 s | GPT-5.6 vs Extend Max Context |
| Qwen 3.8 Flash | 84.1 | $2.98 | 122 s | GPT-5.6 vs Qwen 3.8 Flash |
GPT-5.6 against every other model
- GPT-5.6 vs LlamaParse Extract Agentic Plus
- GPT-5.6 vs LlamaParse Extract Agentic
- GPT-5.6 vs LlamaParse Extract Cost-Effective
- GPT-5.6 vs Reducto Deep Extract
- GPT-5.6 vs Reducto Extract
- GPT-5.6 vs Extend Extract
- GPT-5.6 vs GPT-5.5
- GPT-5.6 vs GPT-5.4 Nano
- GPT-5.6 vs Gemma 4 26B
- GPT-5.6 vs Qwen 3.8 27B
- GPT-5.6 vs Qwen 3.5 35B
- GPT-5.6 vs Qwen 3.5 9B
- GPT-5.6 vs GLM 5.3 Flash
Embed this result
Shows its current rank. Updates with each run.