CITBench - Multi-Step: 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-Step tier: chained operation pipelines). Source: arxiv.org. Saturation forecast: Around December 2026. 12 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Sonnet 4.5 | 43.52 |
| 2 | DeepSeek V3.2 | 41.24 |
| 3 | Gemini 3 Flash | 40.39 |
| 4 | Gemini 3 Pro | 38.13 |
| 5 | GPT-5.2 | 37.05 |
| 6 | Claude Haiku 4.5 | 34.71 |
| 7 | GLM-4.7 | 27.06 |
| 8 | Qwen 3 Coder Flash | 17.9 |
| 9 | Qwen 3 235B A22B | 17.24 |
| 10 | Qwen 3 32B | 15.57 |
| 11 | GPT-4o | 12.93 |
| 12 | Qwen 3 14B | 4.45 |
Interactive version: theaggregate.ai/benchmark?slug=citbench-multi-step · How It Works · Data refreshed daily, snapshot 2026-09-26.