IBA-Bench - Work: leaderboard

Metric: Checkpoint satisfaction (%; share of hidden preference checkpoints satisfied by the final action, work domain; mean of three LLM judges: DeepSeek-V3.2, Qwen3-4B-Instruct and Qwen3-30B-A3B-Instruct; retrieval-augmented baseline with bge-m3 top-3 chunks). Source: arxiv.org. Saturation forecast: Around June 2028. 10 models tracked.

Top models

#ModelScore
1GPT-5 Mini63.9
2GPT-5.162.4
3QwQ-32B61.6
4Qwen 3 4B Instruct60.4
5DeepSeek V3.260.1
6DeepSeek V356.6
7GPT-4o Mini54.9
8GPT-4o54.1
9Qwen 2.5 7B Instruct51
10GLM-4 9B Chat47.3

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