JFinBench - No Tools - Nissho: leaderboard

321examples, 25sections, 6examfamilies; mean3runs. OneAPIcall/question, 131072outputtokens(65536forGemini), nativeeffort/thinking retained. MCQ extraction; GPT5.1 judging freeanswers undernativepartialcreditrubrics; exammean overexamples, overallmeanover6exams. Failure attempts0 pernativepolicy. Public321taskset reconstructablefromofficialsourcePDFrecipes; HFaloneomits57examples/55answers. Metric: Score (%). Source: https://arxiv.org/html/2611.18130v1

Metric: Score (%). Source: arxiv.org. Saturation forecast: Around February 2027. 12 models tracked. Updated 17 Nov 2026.

Top models

#ModelSettingScoreOverall rank
1Claude Fable 5Max100#28
2Gemini 3.8 FlashHigh98.2#39
3GPT-6 AstraMax96.9#8
4Kimi K3.1 ProMax96.8#32
5Kimi K3Max92.6#59
6DeepSeek V4.1 Flash (1017)Max89.9#44
7Qwen 3.8 FlashThinking87.4#128
8Kimi K2.6Thinking84.6#191
9GLM-5.3 FlashMax80.4#136
10MiMo-V2.6-ProThinking76.4#86
11Mistral Large 4Thinking74.8#143
12MiMo-V2.6-FlashThinking65.3#176

No result here: #1 Claude Fable 5.2, #2 GPT-6.2 Astra, #3 Gemini 4 Carbon.

Interactive version: theaggregate.ai/benchmark?slug=jfinbench-no-tools-nissho · How It Works · Data refreshed daily, snapshot 2026-11-17.