Investor's Last Exam · 2026-10-27
Model rankings and capability differences 模型排名与能力差异
Compare overall performance, then examine differences across task categories, evidence depth and scoring dimensions. 先比较综合得分，再从题目类别、证据深度和评分维度了解各模型的优势与短板。
86 tasks · 1472 scored criteria · task-specific information cutoffs · results as of 27 October 2026. 86 道题 · 1472 条计分评分点 · 按题限定信息时点 · 结果截至 2026 年 10 月 27 日。
01 / the score总得分
Overall model scores模型综合得分
The overall score is the equally weighted average of task scores. This evaluation covers 86 equity and sector-basket research tasks. 综合得分按各题得分等权取平均。本次评测覆盖86道股票与行业组合研究题目。
11 fully evaluated models · 86 tasks11 个完整评测模型 · 86 道题
02 / the full result set完整结果
Detailed model results模型详细结果
Scores cover 86 tasks; resource means use each model’s corresponding scored runs. Partial coverage is listed without a rank.得分覆盖 86 题；资源均值来自各模型对应的计分运行。
03 / by task category按题目类别
Seven task families七类投研与行情任务
Compare all fully evaluated models across seven task families.比较全部完整参评模型在七类投研与行情任务中的表现。
04 / category scores分类得分
Rankings by task category各类题目的得分排名
Compare model rankings within each research category.在各研究类别内比较模型得分与排名。
05 / by evidence depth按证据深度拆
Performance across evidence depths不同证据深度下的表现
Compare performance by evidence-document count on the 40 tasks with reference evidence graphs.按40道具备参考证据图的题目的文档数量分组比较表现。
06 / rubric capabilities能力分项
Capabilities, by rubric按 Rubrics 看模型能力
Compare evidence, reasoning, calculations and decisions, or focus on forecast results.比较证据、推理、计算与决策，也可单独查看预测结果。
07 / every cell全矩阵
Model × task category模型 × 题目类别
Scores across seven task families, with the overall score at left.七类任务的得分，左侧为综合得分。