IBA-Bench - Exercise and Health: leaderboard

Metric: Checkpoint satisfaction (%; share of hidden preference checkpoints satisfied by the final action, exercise and health 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 March 2027. 10 models tracked.

Top models

#ModelScore
1GPT-5 Mini78.3
2QwQ-32B75.2
3GPT-5.173.9
4Qwen 3 4B Instruct73.4
5DeepSeek V3.272.2
6DeepSeek V371.4
7GPT-4o69.5
8GPT-4o Mini68.1
9Qwen 2.5 7B Instruct67.1
10GLM-4 9B Chat65.4

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