XpertBench: leaderboard

Expert-level open-ended professional tasks from XpertBench (2026), written by domain experts and graded against 15 to 40 weighted expert rubric checkpoints each, in finance, law, education, engineering and applied sciences, and the humanities and social sciences. Weighted rubric score in percent on the 245-task XpertBench-Gold subset, each checkpoint judged by Gemini 2.5 Pro with an expert-graded exemplar; higher is better.

Metric: Weighted rubric score (%) on the XpertBench-Gold subset (245 expert-written open-ended professional tasks, 15-40 weighted binary checkpoints each); ShotJudge: Gemini 2.5 Pro judges every checkpoint with an expert-graded GPT-5 response as one-shot exemplar; overall over all 245 tasks; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 14 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1Claude Opus 4.6Thinking66.2#113
2GPT-5.4High64.78#135
3Doubao-2.0-pro64.51#149
4GPT-5.2High61.03#185
5GPT-5High58.7#178
6Gemini 3.1 Pro (Preview)55.05#99
7Gemini 3 Pro53.43#145
8Kimi K2 (Thinking)53.3
9Gemini 3 Flash50.98#175
10Claude Sonnet 4.5Thinking46#261
11DeepSeek V3.1 TerminusThinking45.1#437
12GLM-4.644.3#498
13Doubao-Seed-1.6Thinking42.1
14Gemini 2.5 Pro38.2#276

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

Interactive version: theaggregate.ai/benchmark?slug=xpertbench · How It Works · Data refreshed daily, snapshot 2026-11-21.