IBA-Bench - Planning: leaderboard

Metric: Checkpoint satisfaction (%; share of hidden preference checkpoints satisfied by the final action, planning 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 January 2028. 10 models tracked.

Top models

#ModelScore
1GPT-5 Mini69.8
2GPT-5.165.6
3QwQ-32B64.5
4DeepSeek V3.263.2
5Qwen 3 4B Instruct62.8
6DeepSeek V362.7
7GPT-4o57.9
8GPT-4o Mini57.9
9Qwen 2.5 7B Instruct55.5
10GLM-4 9B Chat53.5

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