MosaicLeaks: leaderboard
Multi-hop deep-research questions whose hops alternate between private enterprise documents and a fixed public web corpus; MosaicLeaks (Gurung et al., 2026). Strict chain success in percent over 344 test tasks (every hop answered correctly), mean of three runs with the default agent prompt; higher is better.
Metric: Strict chain success (%; share of 344 test multi-hop research tasks with every hop answered correctly, mean of three runs; default prompt). Source: arxiv.org. Saturation forecast: Around February 2027. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemma 4 31B (IT) | 63.6 | #343 |
| 2 | GPT-OSS-20B | 52.4 | #1150 |
| 3 | Step 3.5 Flash | 51.4 | #508 |
| 4 | Qwen 3 4B Instruct | 48.7 | #1660 |
| 5 | Qwen 3 8B | 48.4 | #1838 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=mosaicleaks · How It Works · Data refreshed daily, snapshot 2026-11-12.