본문으로 건너뛰기
AXyNowAX IS NOW
기술 출력

📄문서·시각자료 이해

Document & visual understanding

채점 권위: ★ 기업 문서·매거진·시각자료 직접 검수

외피 — 산업 도메인
기업 PDF·PPT·논문·인포그래픽·차트·표 이해
내용 — 측정하는 AI 능력
  • · 페이지 이미지에서 표·차트·도표·각주·본문을 결합한 근거 추출
  • · 2열 논문·PPT형 대시보드·인포그래픽의 reading order와 구조 파싱
  • · 로그축·절단축·이중축·외삽·unanswerable 함정 보정

★ 비전 모델 한정 카테고리 — 비전 미지원 모델은 N/A 표시

모델별 종합 점수

✓ 챗봇 1턴

측정일 2026-06-04T07:35:33+00:00 · 5개 항목 × 100점 기준

채점자 editor · max_tokens 32768 · temp 0.7 · attempts 3 · reasoning_effort medium

모델
1ClaudeClaude Opus 4.8
9/9100100100100100100.0
2Google GeminiGemini 3.1 Pro
9/9100100100100100100.0
3Google GeminiGemini 3.5 Flash
9/9100100100100100100.0
4Google GeminiGemini 3.1 Flash Lite
9/9100100100100100100.0
5OpenAIGPT-5.5
9/9100100100100100100.0
6OpenAIGPT-5.4 Mini
9/9100100100100100100.0
7Moonshot AIKimi K2.6
9/9100100100100100100.0
8QWenQwen 3.6 35B A3B
9/9100100100100100100.0
9xAIGrok 4.3
9/9991001009910099.8
10ClaudeClaude Sonnet 4.6
9/9991001001009999.3
11QWenQwen 3.7 Plus
9/996100971009997.8
12MiniMaxMiniMax M3
9/9989798969897.4
13StepFunStep 3.7 Flash
9/9959796969896.3
14QWenQwen 3.5 9B
9/9959193899493.0
15NaverHyperCLOVAX SEED Think 32B
9/99492791009992.8
16Google GeminiGemma 4 31B
9/9889888979391.7
17Mistral AIMistral Small 4
9/9869588959390.2
18LG AIEXAONE 4.5 33B
9/9904484988784.3
19Google GeminiGemma 4 26B A4B
9/9679474917876.9
20Google GeminiGemma 4 12B
9/9477055706157.0

문항별 점수

9 문항

각 문항당 모델 세부 점수. 응답 원문·근거는 문항 카드 우측 링크.

문서·시각자료 이해 · 문항 1기술보고서 1페이지 — 막대 데이터레이블 + 표 조회공개

기술보고서 1페이지 — 막대 데이터레이블 + 표 조회

본문·raw·근거 →
모델
정확성의도 파악신중함한국 맥락짜임새avg
Claude Opus 4.8Anthropic
100100100100100100
Gemini 3.1 ProGoogle
100100100100100100
Gemini 3.5 FlashGoogle
100100100100100100
Gemini 3.1 Flash LiteGoogle
100100100100100100
GPT-5.5OpenAI
100100100100100100
GPT-5.4 MiniOpenAI
100100100100100100
Kimi K2.6Moonshot
100100100100100100
Qwen 3.6 35B A3BAlibaba
100100100100100100
Grok 4.3xAI
100100100100100100
Claude Sonnet 4.6Anthropic
100100100100100100
Qwen 3.7 PlusAlibaba
100100100100100100
MiniMax M3Minimax
100100100100100100
Step 3.7 FlashStepFun
100100100100100100
Qwen 3.5 9BAlibaba
989290929594
HyperCLOVAX SEED Think 32BNaver
100100100100100100
Gemma 4 31BGoogle
100100100100100100
Mistral Small 4Mistral
100100100100100100
EXAONE 4.5 33BLG AI
1006010010010096
Gemma 4 26B A4BGoogle
100100100100100100
Gemma 4 12BGoogle
487070748869
문서·시각자료 이해 · 문항 2경영실적 보고서 — 라인 보간 + 집계표 + 각주 정정비공개

MRR 보간·순증 집계·각주 반영

본문·raw·근거 →
모델
정확성의도 파악신중함한국 맥락짜임새avg
Claude Opus 4.8Anthropic
100100100100100100
Gemini 3.1 ProGoogle
100100100100100100
Gemini 3.5 FlashGoogle
100100100100100100
Gemini 3.1 Flash LiteGoogle
100100100100100100
GPT-5.5OpenAI
100100100100100100
GPT-5.4 MiniOpenAI
100100100100100100
Kimi K2.6Moonshot
100100100100100100
Qwen 3.6 35B A3BAlibaba
100100100100100100
Grok 4.3xAI
100100100100100100
Claude Sonnet 4.6Anthropic
100100100100100100
Qwen 3.7 PlusAlibaba
9610010010010099
MiniMax M3Minimax
100981009610099
Step 3.7 FlashStepFun
9698100969898
Qwen 3.5 9BAlibaba
929093909292
HyperCLOVAX SEED Think 32BNaver
1009010010010099
Gemma 4 31BGoogle
8510010010010096
Mistral Small 4Mistral
100100100100100100
EXAONE 4.5 33BLG AI
1006010010010096
Gemma 4 26B A4BGoogle
20100809010071
Gemma 4 12BGoogle
256545707552
문서·시각자료 이해 · 문항 3IR 재무자료 — 중첩헤더 재무표 + 비율 산출 + 일회성 제외비공개

구독 비중·영업이익률·경상 매출총이익률

본문·raw·근거 →
모델
정확성의도 파악신중함한국 맥락짜임새avg
Claude Opus 4.8Anthropic
100100100100100100
Gemini 3.1 ProGoogle
100100100100100100
Gemini 3.5 FlashGoogle
100100100100100100
Gemini 3.1 Flash LiteGoogle
100100100100100100
GPT-5.5OpenAI
100100100100100100
GPT-5.4 MiniOpenAI
100100100100100100
Kimi K2.6Moonshot
100100100100100100
Qwen 3.6 35B A3BAlibaba
100100100100100100
Grok 4.3xAI
951001009010098
Claude Sonnet 4.6Anthropic
100100100100100100
Qwen 3.7 PlusAlibaba
100100100100100100
MiniMax M3Minimax
100961009810099
Step 3.7 FlashStepFun
100961009810099
Qwen 3.5 9BAlibaba
969092929393
HyperCLOVAX SEED Think 32BNaver
1009010010010099
Gemma 4 31BGoogle
100100100100100100
Mistral Small 4Mistral
100100100100100100
EXAONE 4.5 33BLG AI
1006010010010096
Gemma 4 26B A4BGoogle
100100100100100100
Gemma 4 12BGoogle
908084858686
문서·시각자료 이해 · 문항 4제품 스펙시트 — 병합셀 표 다중조건 필터비공개

조건 충족 모델명 하나만 출력

본문·raw·근거 →
모델
정확성의도 파악신중함한국 맥락짜임새avg
Claude Opus 4.8Anthropic
100100100100100100
Gemini 3.1 ProGoogle
100100100100100100
Gemini 3.5 FlashGoogle
100100100100100100
Gemini 3.1 Flash LiteGoogle
100100100100100100
GPT-5.5OpenAI
100100100100100100
GPT-5.4 MiniOpenAI
100100100100100100
Kimi K2.6Moonshot
100100100100100100
Qwen 3.6 35B A3BAlibaba
100100100100100100
Grok 4.3xAI
100100100100100100
Claude Sonnet 4.6Anthropic
100100100100100100
Qwen 3.7 PlusAlibaba
100100100100100100
MiniMax M3Minimax
10010010098100100
Step 3.7 FlashStepFun
10010010098100100
Qwen 3.5 9BAlibaba
959888889593
HyperCLOVAX SEED Think 32BNaver
100100100100100100
Gemma 4 31BGoogle
100100100100100100
Mistral Small 4Mistral
100100100100100100
EXAONE 4.5 33BLG AI
1004010010010094
Gemma 4 26B A4BGoogle
100100100100100100
Gemma 4 12BGoogle
358550703045
문서·시각자료 이해 · 문항 5시장 포지셔닝 브리프 — 절단축 차트 + unanswerable비공개

절단 y축·점유율 격차·없는 D사 정보 거부

본문·raw·근거 →
모델
정확성의도 파악신중함한국 맥락짜임새avg
Claude Opus 4.8Anthropic
100100100100100100
Gemini 3.1 ProGoogle
100100100100100100
Gemini 3.5 FlashGoogle
100100100100100100
Gemini 3.1 Flash LiteGoogle
100100100100100100
GPT-5.5OpenAI
100100100100100100
GPT-5.4 MiniOpenAI
100100100100100100
Kimi K2.6Moonshot
100100100100100100
Qwen 3.6 35B A3BAlibaba
100100100100100100
Grok 4.3xAI
100100100100100100
Claude Sonnet 4.6Anthropic
100100100100100100
Qwen 3.7 PlusAlibaba
1001009210010098
MiniMax M3Minimax
989597929596
Step 3.7 FlashStepFun
969695929695
Qwen 3.5 9BAlibaba
959095829092
HyperCLOVAX SEED Think 32BNaver
90904010010084
Gemma 4 31BGoogle
9010010010010097
Mistral Small 4Mistral
90901008510094
EXAONE 4.5 33BLG AI
904010010010091
Gemma 4 26B A4BGoogle
9010010010010097
Gemma 4 12BGoogle
807882787879
문서·시각자료 이해 · 문항 6용량·비용 최적화 차트 — U자 최저점 + 적분 + SLA비공개

U자 비용곡선·사다리꼴 적분·이중축 함정

본문·raw·근거 →
모델
정확성의도 파악신중함한국 맥락짜임새avg
Claude Opus 4.8Anthropic
100100100100100100
Gemini 3.1 ProGoogle
100100100100100100
Gemini 3.5 FlashGoogle
100100100100100100
Gemini 3.1 Flash LiteGoogle
100100100100100100
GPT-5.5OpenAI
100100100100100100
GPT-5.4 MiniOpenAI
100100100100100100
Kimi K2.6Moonshot
100100100100100100
Qwen 3.6 35B A3BAlibaba
100100100100100100
Grok 4.3xAI
100100100100100100
Claude Sonnet 4.6Anthropic
100100100100100100
Qwen 3.7 PlusAlibaba
100100100100100100
MiniMax M3Minimax
100981009610099
Step 3.7 FlashStepFun
100981009610099
Qwen 3.5 9BAlibaba
969294909594
HyperCLOVAX SEED Think 32BNaver
1009010010010099
Gemma 4 31BGoogle
100100100100100100
Mistral Small 4Mistral
100100100100100100
EXAONE 4.5 33BLG AI
100401001009091
Gemma 4 26B A4BGoogle
100100100100100100
Gemma 4 12BGoogle
507050728063
문서·시각자료 이해 · 문항 72열 기술논문 — 수식·표·그림 3-hop cross-reference비공개

2열 논문 reading order·수식 대입·SLA 판독

본문·raw·근거 →
모델
정확성의도 파악신중함한국 맥락짜임새avg
Claude Opus 4.8Anthropic
100100100100100100
Gemini 3.1 ProGoogle
100100100100100100
Gemini 3.5 FlashGoogle
100100100100100100
Gemini 3.1 Flash LiteGoogle
100100100100100100
GPT-5.5OpenAI
100100100100100100
GPT-5.4 MiniOpenAI
100100100100100100
Kimi K2.6Moonshot
100100100100100100
Qwen 3.6 35B A3BAlibaba
100100100100100100
Grok 4.3xAI
100100100100100100
Claude Sonnet 4.6Anthropic
100100100100100100
Qwen 3.7 PlusAlibaba
961001001009698
MiniMax M3Minimax
100961009810099
Step 3.7 FlashStepFun
100961009810099
Qwen 3.5 9BAlibaba
959095929594
HyperCLOVAX SEED Think 32BNaver
85905010010084
Gemma 4 31BGoogle
559040907064
Mistral Small 4Mistral
8085100909088
EXAONE 4.5 33BLG AI
702030905053
Gemma 4 26B A4BGoogle
15800702026
Gemma 4 12BGoogle
255825603033
문서·시각자료 이해 · 문항 82열 그로스 논문 — 로그스케일 + 히트맵 + 코호트 표비공개

반로그 축 오독저항·히트맵 2D 셀 판독

본문·raw·근거 →
모델
정확성의도 파악신중함한국 맥락짜임새avg
Claude Opus 4.8Anthropic
100100100100100100
Gemini 3.1 ProGoogle
100100100100100100
Gemini 3.5 FlashGoogle
100100100100100100
Gemini 3.1 Flash LiteGoogle
100100100100100100
GPT-5.5OpenAI
100100100100100100
GPT-5.4 MiniOpenAI
100100100100100100
Kimi K2.6Moonshot
100100100100100100
Qwen 3.6 35B A3BAlibaba
100100100100100100
Grok 4.3xAI
100100100100100100
Claude Sonnet 4.6Anthropic
901001001009094
Qwen 3.7 PlusAlibaba
74100781009285
MiniMax M3Minimax
809282909086
Step 3.7 FlashStepFun
649070889078
Qwen 3.5 9BAlibaba
908893829290
HyperCLOVAX SEED Think 32BNaver
7090201009071
Gemma 4 31BGoogle
659050857068
Mistral Small 4Mistral
409020906052
EXAONE 4.5 33BLG AI
502030904546
Gemma 4 26B A4BGoogle
15800702026
Gemma 4 12BGoogle
286035603036
문서·시각자료 이해 · 문항 92열 스케일링 논문 — 3차 변곡 + 로그-로그 멱법칙 + 적분비공개

비선형 함수 정량추론 극단

본문·raw·근거 →
모델
정확성의도 파악신중함한국 맥락짜임새avg
Claude Opus 4.8Anthropic
100100100100100100
Gemini 3.1 ProGoogle
100100100100100100
Gemini 3.5 FlashGoogle
100100100100100100
Gemini 3.1 Flash LiteGoogle
100100100100100100
GPT-5.5OpenAI
100100100100100100
GPT-5.4 MiniOpenAI
100100100100100100
Kimi K2.6Moonshot
100100100100100100
Qwen 3.6 35B A3BAlibaba
100100100100100100
Grok 4.3xAI
100100100100100100
Claude Sonnet 4.6Anthropic
100100100100100100
Qwen 3.7 PlusAlibaba
100100100100100100
MiniMax M3Minimax
100961009610099
Step 3.7 FlashStepFun
100961009610099
Qwen 3.5 9BAlibaba
979295929695
HyperCLOVAX SEED Think 32BNaver
1009010010010099
Gemma 4 31BGoogle
10010010095100100
Mistral Small 4Mistral
659070909078
EXAONE 4.5 33BLG AI
1006010010010096
Gemma 4 26B A4BGoogle
658590856072
Gemma 4 12BGoogle
386055625050