개인 의사결정EP5

🚗차사고·자동차 분쟁

Car accident·dispute

외피 — 산업 도메인

차사고·보험 분쟁

내용 — 측정하는 AI 능력

· 과실비율 추론 (도로교통법)
· 보험·구상 절차
· 형사·민사 분기

모델별 종합 점수

✓ 챗봇 1턴

측정일 2026-06-05T02:40:53+00:00 · 5개 항목 × 100점 기준

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

모델
1Gemini 3.1 Pro	5/5	92	90	90	93	90	91.3
2Claude Opus 4.8	5/5	88	90	90	90	91	89.8
3GPT-5.5	5/5	90	88	88	90	91	89.4
4Gemini 3.5 Flash	5/5	89	88	86	90	90	88.8
5DeepSeek V4 Pro	5/5	89	88	87	89	90	88.5
6Claude Sonnet 4.6	5/5	88	87	85	87	90	87.3
7Qwen 3.7 Max	5/5	85	83	82	84	86	84.0
8Kimi K2.6	5/5	81	80	78	78	82	79.6
9Gemma 4 12B	5/5	78	80	78	76	82	78.0
10Gemma 4 31B	5/5	78	77	76	77	80	77.5
11GPT-5.4 Mini	5/5	79	78	76	76	79	77.3
12Gemma 4 26B A4B	5/5	78	77	75	75	80	76.7
13MiniMax M3	5/5	66	87	85	65	92	76.4
14Gemini 3.1 Flash Lite	5/5	76	76	74	76	79	76.0
15Mimo V2.5 Pro	5/5	76	76	74	74	81	75.8
16Qwen 3.7 Plus	5/5	71	82	73	69	85	74.6
17GLM 5.1	5/5	75	77	72	72	80	74.4
18DeepSeek V4 Flash	5/5	74	74	71	71	77	73.1
19Step 3.7 Flash	5/5	63	88	71	63	89	72.2
20Grok 4.3	5/5	73	73	70	69	76	71.6
21Nemotron 3 Ultra 550B	5/5	63	82	63	60	88	68.4
22Qwen 3.6 27B	5/5	57	76	79	55	88	68.0
23Solar Pro 3	5/5	58	71	58	50	80	60.8
24EXAONE 4.5 33B	5/5	54	69	55	49	77	58.4
25HyperCLOVAX SEED Think 32B	5/5	48	64	48	52	80	56.4
26Qwen 3.5 9B	5/5	42	68	55	35	78	52.2
27Qwen 3.6 35B A3B	5/5	52	53	48	46	61	51.0
28Mistral Small 4	5/5	51	55	47	40	60	48.8
29Gemma 4 E2B	5/5	36	47	34	34	45	38.4
30Kanana 2 30B-A3B Thinking	5/5	30	44	37	31	45	36.2
31HyperCLOVAX SEED 1.5B	5/5	33	45	32	32	43	35.6
32LFM2.5 8B-A1B	5/5	27	42	25	24	40	30.0

문항별 점수

5 문항

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

차사고·자동차 분쟁 · 문항 1고속도로 추돌 가해자 + 12주 진단서 대응공개

고속도로 추돌 가해자 + 12주 진단서 대응

본문·raw·근거 →

모델

정확성의도 파악신중함한국 맥락짜임새avg

Gemini 3.1 ProGoogle

929090929091

Claude Opus 4.8Anthropic

889090889089

GPT-5.5OpenAI

888885889088

Gemini 3.5 FlashGoogle

888885888887

DeepSeek V4 ProDeepSeek

908888908889

Claude Sonnet 4.6Anthropic

888885888887

Qwen 3.7 MaxAlibaba

858582858584

Kimi K2.6Moonshot

808078788279

Gemma 4 12BGoogle

788078788479

Gemma 4 31BGoogle

807878808280

GPT-5.4 MiniOpenAI

828078808080

Gemma 4 26B A4BGoogle

787875758077

MiniMax M3Minimax

508282459266

Gemini 3.1 Flash LiteGoogle

757572727874

Mimo V2.5 ProXiaomi

787875758077

Qwen 3.7 PlusAlibaba

828482828483

GLM 5.1Z.ai

787875758077

DeepSeek V4 FlashDeepSeek

757572727874

Step 3.7 FlashStepFun

508555488861

Grok 4.3xAI

757572707573

Nemotron 3 Ultra 550BNVIDIA

568058508863

Qwen 3.6 27BAlibaba

557075508564

Solar Pro 3Upstage

607260528263

EXAONE 4.5 33BLG AI

607060557863

HyperCLOVAX SEED Think 32BNaver

406040408049

Qwen 3.5 9BAlibaba

407055327851

Qwen 3.6 35B A3BAlibaba

656562607064

Mistral Small 4Mistral

606055506557

Gemma 4 E2BGoogle

324530284334

Kanana 2 30B-A3B ThinkingKakao

405845405546

HyperCLOVAX SEED 1.5BNaver

274125253930

LFM2.5 8B-A1BLiquid AI

294327274132

차사고·자동차 분쟁 · 문항 2자차보험 청구 vs 대물 청구 차이비공개

자차·대물·대인 청구 트리

본문·raw·근거 →

모델

정확성의도 파악신중함한국 맥락짜임새avg

Gemini 3.1 ProGoogle

908888908889

Claude Opus 4.8Anthropic

889290929291

GPT-5.5OpenAI

888888889088

Gemini 3.5 FlashGoogle

888885909088

DeepSeek V4 ProDeepSeek

858582858885

Claude Sonnet 4.6Anthropic

888885859087

Qwen 3.7 MaxAlibaba

808078808280

Kimi K2.6Moonshot

787875758077

Gemma 4 12BGoogle

747874728075

Gemma 4 31BGoogle

757572727874

GPT-5.4 MiniOpenAI

787875758077

Gemma 4 26B A4BGoogle

757572727874

MiniMax M3Minimax

809090859287

Gemini 3.1 Flash LiteGoogle

727270707571

Mimo V2.5 ProXiaomi

787875758277

Qwen 3.7 PlusAlibaba

788484748680

GLM 5.1Z.ai

757572727874

DeepSeek V4 FlashDeepSeek

757572727874

Step 3.7 FlashStepFun

729080759080

Grok 4.3xAI

727270687571

Nemotron 3 Ultra 550BNVIDIA

708678748878

Qwen 3.6 27BAlibaba

657580709075

Solar Pro 3Upstage

607260528062

EXAONE 4.5 33BLG AI

557058507559

HyperCLOVAX SEED Think 32BNaver

608060608066

Qwen 3.5 9BAlibaba

628072608269

Qwen 3.6 35B A3BAlibaba

606058556559

Mistral Small 4Mistral

505548406049

Gemma 4 E2BGoogle

405038384842

Kanana 2 30B-A3B ThinkingKakao

526852507257

HyperCLOVAX SEED 1.5BNaver

394938384741

LFM2.5 8B-A1BLiquid AI

314529294434

차사고·자동차 분쟁 · 문항 3음주 운전 적발 시 행정·형사·보험 영향비공개

음주 0.03~0.08~0.2 단계별 처벌

본문·raw·근거 →

모델

정확성의도 파악신중함한국 맥락짜임새avg

Gemini 3.1 ProGoogle

929088929291

Claude Opus 4.8Anthropic

889088909089

GPT-5.5OpenAI

908888929290

Gemini 3.5 FlashGoogle

888885908888

DeepSeek V4 ProDeepSeek

888885889088

Claude Sonnet 4.6Anthropic

888885909088

Qwen 3.7 MaxAlibaba

828078828581

Kimi K2.6Moonshot

858282828583

Gemma 4 12BGoogle

788078788279

Gemma 4 31BGoogle

787875788078

GPT-5.4 MiniOpenAI

787875757876

Gemma 4 26B A4BGoogle

807878808280

MiniMax M3Minimax

658575709276

Gemini 3.1 Flash LiteGoogle

727270707571

Mimo V2.5 ProXiaomi

727270707872

Qwen 3.7 PlusAlibaba

748274788478

GLM 5.1Z.ai

787875788078

DeepSeek V4 FlashDeepSeek

727270707571

Step 3.7 FlashStepFun

628878608872

Grok 4.3xAI

707268687570

Nemotron 3 Ultra 550BNVIDIA

768470708876

Qwen 3.6 27BAlibaba

707580709076

Solar Pro 3Upstage

607260558264

EXAONE 4.5 33BLG AI

556855507859

HyperCLOVAX SEED Think 32BNaver

404060408050

Qwen 3.5 9BAlibaba

385858327549

Qwen 3.6 35B A3BAlibaba

555552506054

Mistral Small 4Mistral

505545356047

Gemma 4 E2BGoogle

344532324336

Kanana 2 30B-A3B ThinkingKakao

254035254532

HyperCLOVAX SEED 1.5BNaver

304329294133

LFM2.5 8B-A1BLiquid AI

233821193826

차사고·자동차 분쟁 · 문항 4주차장 사고 — CCTV·블랙박스 증거 활용비공개

주차장 사고 책임 입증

본문·raw·근거 →

모델

정확성의도 파악신중함한국 맥락짜임새avg

Gemini 3.1 ProGoogle

929090959092

Claude Opus 4.8Anthropic

869092889089

GPT-5.5OpenAI

908888889289

Gemini 3.5 FlashGoogle

908888929090

DeepSeek V4 ProDeepSeek

909088909290

Claude Sonnet 4.6Anthropic

858582859085

Qwen 3.7 MaxAlibaba

858280858884

Kimi K2.6Moonshot

787875758077

Gemma 4 12BGoogle

808080808280

Gemma 4 31BGoogle

807878808280

GPT-5.4 MiniOpenAI

787875757876

Gemma 4 26B A4BGoogle

807878788279

MiniMax M3Minimax

709088659078

Gemini 3.1 Flash LiteGoogle

787878788078

Mimo V2.5 ProXiaomi

757575758276

Qwen 3.7 PlusAlibaba

487838428254

GLM 5.1Z.ai

607055507560

DeepSeek V4 FlashDeepSeek

707268657569

Step 3.7 FlashStepFun

628862608869

Grok 4.3xAI

727268707571

Nemotron 3 Ultra 550BNVIDIA

427232408450

Qwen 3.6 27BAlibaba

407575458560

Solar Pro 3Upstage

506852427855

EXAONE 4.5 33BLG AI

456548427552

HyperCLOVAX SEED Think 32BNaver

406020608051

Qwen 3.5 9BAlibaba

185520157532

Qwen 3.6 35B A3BAlibaba

253020205027

Mistral Small 4Mistral

455040355543

Gemma 4 E2BGoogle

384836364640

Kanana 2 30B-A3B ThinkingKakao

254035254532

HyperCLOVAX SEED 1.5BNaver

374736364539

LFM2.5 8B-A1BLiquid AI

254123213928

차사고·자동차 분쟁 · 문항 5보험금 사기 의심 시 대응 (작은 사고 큰 청구)비공개

사기 의심 시 보험사 신고 동선

본문·raw·근거 →

모델

정확성의도 파악신중함한국 맥락짜임새avg

Gemini 3.1 ProGoogle

959292959294

Claude Opus 4.8Anthropic

909092929291

GPT-5.5OpenAI

929092929292

Gemini 3.5 FlashGoogle

908888929290

DeepSeek V4 ProDeepSeek

929090929291

Claude Sonnet 4.6Anthropic

908888889289

Qwen 3.7 MaxAlibaba

928890909090

Kimi K2.6Moonshot

828280828582

Gemma 4 12BGoogle

788078728477

Gemma 4 31BGoogle

787875758077

GPT-5.4 MiniOpenAI

787875757876

Gemma 4 26B A4BGoogle

757572728074

MiniMax M3Minimax

658888589275

Gemini 3.1 Flash LiteGoogle

858282888585

Mimo V2.5 ProXiaomi

787875758277

Qwen 3.7 PlusAlibaba

748486688878

GLM 5.1Z.ai

858282858584

DeepSeek V4 FlashDeepSeek

787875758077

Step 3.7 FlashStepFun

709080749079

Grok 4.3xAI

757570707873

Nemotron 3 Ultra 550BNVIDIA

708676649075

Qwen 3.6 27BAlibaba

558584389065

Solar Pro 3Upstage

587056508060

EXAONE 4.5 33BLG AI

557055487859

HyperCLOVAX SEED Think 32BNaver

608060608066

Qwen 3.5 9BAlibaba

527870388060

Qwen 3.6 35B A3BAlibaba

555550456052

Mistral Small 4Mistral

505545406048

Gemma 4 E2BGoogle

384836364640

Kanana 2 30B-A3B ThinkingKakao

101520151014

HyperCLOVAX SEED 1.5BNaver

324431314235

LFM2.5 8B-A1BLiquid AI

274226244030