Skip to content

Anthropic · Vision LLM

Claude Opus 4.8 on ResumeBench: 85.5 F1, rank 4 of 18

Claude Opus 4.8 ranks 4th of 18 on ResumeBench: 85.5 F1, $90 per 1,000 pages, 31 seconds per resume. It is 1.7 points behind LlamaParse Extract Agentic Plus. LlamaParse Extract Agentic Plus scores 87.2 for $54.

1st of 11 vision LLMs. Strongest on hidden text & prompt injections (+6.3 points), image only (+5.7 points) and scanned & rotated (+4.6 points). Weakest on content as images (−4.6 points), wording traps (−2.6 points) and distractor layouts (−2.5 points).

Rank
4/18
F1
85.5
Precision
86.8
Recall
87.5
Perfect resumes
31/300
Per 1,000 pages
$90
Median per resume
31 s

Averaged over 300 resumes. Dates (scored separately): 97.7. 90% finish within 52 s.

Where Claude Opus 4.8 wins and loses

F1 by kind of resume, against the median system.

Reading the page

Claude Opus 4.8 by reading the page
Kind of resume# of resumesF1MedianGapRank
Handwritten954.751.5+3.33rd
Image only2877.071.4+5.73rd
Sidebar or grid8284.281.9+2.37th
Photo4586.384.0+2.24th

Distractors

Claude Opus 4.8 by distractors
Kind of resume# of resumesF1MedianGapRank
Rotated & warped pages2187.482.9+4.48th
Hidden text & prompt injections1490.884.5+6.33rd
Content as images883.688.2−4.614th
Wording traps1681.984.5−2.612th
Typos kept on purpose3785.283.7+1.68th

Extraction challenges

Claude Opus 4.8 by extraction challenges
Kind of resume# of resumesF1MedianGapRank
Many entries8383.781.9+1.87th
Long bullet lists10183.181.4+1.77th
Split across pages8584.682.2+2.44th
Big skills list6084.883.6+1.27th
Custom sections10583.480.8+2.63rd
Combined headings3785.182.7+2.51st
Promotions8885.583.5+2.06th
Partial dates13884.583.4+1.17th
Overlapping jobs6983.581.7+1.76th
Accented text1385.685.5+0.19th

Noise

Claude Opus 4.8 by noise
Kind of resume# of resumesF1MedianGapRank
Repeated header6787.685.2+2.42nd
References5587.685.2+2.56th
Busy page1880.880.1+0.77th

Source

Claude Opus 4.8 by source
Kind of resume# of resumesF1MedianGapRank
Everyday resumes21286.885.1+1.65th
Expert-built (Canva / LaTeX)3286.183.6+2.52nd
Resume-builder templates1085.481.2+4.24th
Scanned & rotated1987.582.9+4.67th
Distractor layouts1882.885.3−2.512th

Layout

Claude Opus 4.8 by layout
Kind of resume# of resumesF1MedianGapRank
Single column19585.684.0+1.62nd
Two-column2389.786.8+2.83rd
Two-column sidebar5481.980.0+1.97th
Three-column2086.284.8+1.42nd
Table grid895.194.1+1.16th

Length

Claude Opus 4.8 by length
Kind of resume# of resumesF1MedianGapRank
Short14888.387.1+1.15th
Medium9085.884.7+1.23rd
Long6278.474.6+3.83rd

By section of the resume

F1 by resume section, against the median system.

Claude Opus 4.8 by resume section
SectionResumesF1MedianGap
Contact details30093.793.5+0.2
Experience30088.286.3+1.9
Education29790.289.4+0.8
Summary29089.781.7+7.9
Certifications23985.481.1+4.3
Awards10672.975.4−2.4
Skills28395.292.1+3.1
Languages16785.884.3+1.5
Custom sections21851.649.9+1.7
Interests6393.791.7+1.9
Volunteering7884.684.5+0.1
Profiles3561.466.7−5.3
Projects7882.183.2−1.1
Publications5281.878.6+3.2

On the published resumes

Best and worst of the 20 public resumes.

Claude Opus 4.8 on published resumes
ResumeWhat makes it hardF1
Senior Manufacturing EngineerThird-person prose, no bullets93.0
Head of Developer Experience (Internal title: Jedi Master of Analytics) Unusual job titles88.8
Quality Systems LeadPhoto of a printout87.3
Senior Game DesignerGame designer's Canva resume64.6
Senior ocularist and practice ownerHandwritten resume, six pages49.7
senior Consultant - package ImplementationHandwritten resume49.4

How it was run

Anthropic API, Claude Opus 4.8, page images in, JSON schema out. Same PDF, schema and instructions as every other system, one run per resume (4 November 2026).

Claude Opus 4.8 on Anthropic’s site

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 test

Closest alternatives

Nearest in F1.

Embed this result

Shows its current rank. Updates with each run.

Claude Opus 4.8: #4 on ResumeBench resume extraction benchmark
<a href="https://www.resumebench.ai/systems/claude-opus-4-8"><img src="https://www.resumebench.ai/badge/claude-opus-4-8.svg" alt="Claude Opus 4.8: #4 on ResumeBench resume extraction benchmark" height="28"></a>

Choosing a resume extraction model

All 18 models · Leaderboard

Frequently asked questions (FAQs)

How accurate is Claude Opus 4.8 at resume extraction?

85.5 F1 on ResumeBench's 300 resumes, 4th of 18 models.

How much does Claude Opus 4.8 cost to parse resumes?

$90 per 1,000 pages, about $0.251 per resume.

How fast is Claude Opus 4.8?

31 seconds per resume (median).

Which resumes is Claude Opus 4.8 best at?

Hidden text & prompt injections, image only and scanned & rotated. Weakest on content as images, wording traps and distractor layouts.

What is a cheaper alternative to Claude Opus 4.8?

LlamaParse Extract Agentic Plus: 87.2 F1 for $54 per 1,000 pages.