CITBench - Multi-Input: leaderboard

Metric: Average performance (%; mean Level-Accuracy with partial credit, each of repeated runs scored 0 to 1, on the offline single-turn tabular data processing tasks of CITBench, a balanced 50% slice of its 1,296 instances in 18 matching, cleaning, augmentation and transformation task types; temperature 1.0, 8,192 output tokens; Multi-Input tier: several input tables). Source: arxiv.org. Saturation forecast: Around December 2026. 12 models tracked.

Top models

#ModelScore
1Gemini 3 Pro62.63
2Claude Sonnet 4.562.5
3Gemini 3 Flash60.16
4GPT-5.257.7
5GLM-4.756.04
6GPT-4o49.62
7Claude Haiku 4.549.04
8Qwen 3 32B41.38
9Qwen 3 Coder Flash39.76
10Qwen 3 235B A22B38.06
11Qwen 3 14B36.86
12DeepSeek V3.230.04

Interactive version: theaggregate.ai/benchmark?slug=citbench-multi-input · How It Works · Data refreshed daily, snapshot 2026-09-26.