JFinBench - No Tools - FP: 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
2GPT-6 AstraMax99.5#8
3Gemini 3.8 FlashHigh96.9#39
4Kimi K3.1 ProMax87.7#32
5Kimi K3Max81#59
6DeepSeek V4.1 Flash (1017)Low67.7#44
7MiMo-V2.6-ProThinking50.3#86
8Mistral Large 4Thinking44.1#143
9Kimi K2.6Thinking42.1#191
10GLM-5.3 FlashLow41#136
11Qwen 3.8 FlashThinking39.5#128
12MiMo-V2.6-FlashThinking36.4#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-fp · How It Works · Data refreshed daily, snapshot 2026-11-17.