Skip to content
vs

Claude Opus 4.8 vs GPT-5.6: which is better at resume extraction?

Claude Opus 4.8 leads GPT-5.6 by 0.8 points (85.5 vs 84.7 F1). Cost per 1,000 pages: $90 vs $74. Median time per resume: 31 s vs 38 s.

Claude Opus 4.8 wins on 134 of 300 resumes, GPT-5.6 on 122, 44 tied.

85.51

Claude Opus 4.8

A · Vision LLM

+0.76A − B, points of F1

84.75

GPT-5.6

B · Vision LLM

A better on 134level on 44B better on 122

Claude Opus 4.8 is ahead by 0.76 points on these 300 resumes.

  • Many entries 83A 83.7 · B 84.5−0.8
  • Long bullet lists 101A 83.1 · B 84.9−1.8
  • Split across pages 85A 84.6 · B 84.9−0.3
  • Big skills list 60A 84.8 · B 85.2−0.3
  • Custom sections 105A 83.4 · B 82.2+1.2
  • Combined headings 37A 85.1 · B 84.1+1.1
  • Promotions 88A 85.5 · B 85.0+0.5
  • Partial dates 138A 84.5 · B 85.7−1.2
  • Overlapping jobs 69A 83.5 · B 83.2+0.3
  • Accented text 13A 85.6 · B 86.5−1.0

Side by side

Claude Opus 4.8 vs GPT-5.6
Claude Opus 4.8GPT-5.6
TypeVision LLMVision LLM
MakerAnthropicOpenAI
Rank4 of 186 of 18
F185.584.7
Precision86.885.9
Recall87.587.2
Cost per 1,000 pages$90$74
Median time per resume31 s38 s
Resumes won134122

Claude Opus 4.8 in detailGPT-5.6 in detail

Where each one wins

Where the two differ most.

Claude Opus 4.8 vs GPT-5.6 by kind of resume
Kind of resume# of resumesClaude Opus 4.8GPT-5.6Gap
Scanned & rotated (source)1987.571.6+16.0
Rotated & warped pages (distractors)2187.473.2+14.1
Hidden text & prompt injections (distractors)1490.877.5+13.3
Image only (reading the page)2877.065.4+11.6
Content as images (distractors)883.693.0−9.4
Distractor layouts (source)1882.890.1−7.3
Wording traps (distractors)1681.989.1−7.2
Typos kept on purpose (distractors)3785.280.6+4.7
Two-column sidebar (layout)5481.978.3+3.5
Busy page (noise)1880.884.0−3.2
Sidebar or grid (reading the page)8284.281.7+2.5
Long bullet lists (extraction challenges)10183.184.9−1.8
Claude Opus 4.8 vs GPT-5.6 by section
SectionResumesClaude Opus 4.8GPT-5.6Gap
Contact details30093.794.1−0.3
Experience30088.288.4−0.2
Education29790.291.9−1.7
Summary29089.783.4+6.2
Certifications23985.477.9+7.5
Awards10672.981.2−8.2
Skills28395.293.2+2.0
Languages16785.885.9−0.0
Custom sections21851.651.2+0.4
Interests6393.791.4+2.3
Volunteering7884.685.7−1.1
Profiles3561.469.4−7.9
Projects7882.188.2−6.1
Publications5281.877.8+4.0

Published resumes they disagree on most

Open one to see both outputs field by field.

Which to pick

Claude Opus 4.8 is more accurate. GPT-5.6 costs 18% less for 0.8 fewer points.

Claude Opus 4.8 does better on scanned & rotated, rotated & warped pages and hidden text & prompt injections. GPT-5.6 does better on content as images, distractor layouts and wording traps.

Claude Opus 4.8 against every other model

GPT-5.6 against every other model

Popular comparisons

Frequently asked questions (FAQs)

Which is better for resume extraction, Claude Opus 4.8 or GPT-5.6?

Claude Opus 4.8: 85.5 F1 against 84.7.

Which is cheaper, Claude Opus 4.8 or GPT-5.6?

GPT-5.6: $74 per 1,000 pages against $90.

Where does each one do better?

Claude Opus 4.8: scanned & rotated, rotated & warped pages and hidden text & prompt injections. GPT-5.6: content as images, distractor layouts and wording traps.