AISE-Bench: leaderboard
Academic information-seeking agent benchmark: models plan multi-step API calls over a scholarly knowledge graph to answer research queries; F1-LM (%) combines answer correctness and completeness with faithfulness to the API results actually returned.
Metric: F1-LM (%). Source: aise-bench.github.io. Saturation forecast: Around September 2027. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro | 56.06 | #83 |
| 2 | Qwen 3 235B A22B | 48.56 | #326 |
| 3 | DeepSeek V3.2 | 46.49 | #211 |
| 4 | GPT-5.2 | 44.63 | #114 |
| 5 | GLM-4.7 | 36.15 | #202 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #6 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=aise-bench · How It Works · Data refreshed daily, snapshot 2026-11-06.