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
| # | Model | Score |
|---|---|---|
| 1 | GPT-5 Mini | 63.9 |
| 2 | GPT-5.1 | 62.4 |
| 3 | QwQ-32B | 61.6 |
| 4 | Qwen 3 4B Instruct | 60.4 |
| 5 | DeepSeek V3.2 | 60.1 |
| 6 | DeepSeek V3 | 56.6 |
| 7 | GPT-4o Mini | 54.9 |
| 8 | GPT-4o | 54.1 |
| 9 | Qwen 2.5 7B Instruct | 51 |
| 10 | GLM-4 9B Chat | 47.3 |
Interactive version: theaggregate.ai/benchmark?slug=iba-bench-work · How It Works · Data refreshed daily, snapshot 2026-09-26.