IBA-Bench - Daily Consumption: leaderboard

Metric: Checkpoint satisfaction (%; share of hidden preference checkpoints satisfied by the final action, daily consumption 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 December 2026. 10 models tracked.

Top models

#ModelScore
1QwQ-32B80.6
2GPT-5.179.9
3DeepSeek V3.279.8
4Qwen 3 4B Instruct77.8
5GPT-4o Mini77.4
6Qwen 2.5 7B Instruct77.3
7DeepSeek V377.1
8GPT-5 Mini76.3
9GPT-4o75.6
10GLM-4 9B Chat70.7

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