JFinBench - No Tools - Zensho: 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 5Max98#28
2GPT-6 AstraMax96#8
3Kimi K3Max92.9#59
4Kimi K3.1 ProMax92.9#32
5DeepSeek V4.1 Flash (1017)Max88.9#44
6Gemini 3.8 FlashHigh85.9#39
7MiMo-V2.6-ProThinking82.8#86
8GLM-5.3 FlashMax79.8#136
9Qwen 3.8 FlashThinking79.8#128
10Kimi K2.6Thinking73.7#191
11Mistral Large 4Thinking73.7#143
12MiMo-V2.6-FlashThinking72.7#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-zensho · How It Works · Data refreshed daily, snapshot 2026-11-17.