AIR-Bench Long-Doc: leaderboard
Metric: Long-document Recall@10 (%). Source: huggingface.co. 45 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | BM25 (Reranker: mmarco-mMiniLMv2-L12-H384-v1) | 75.01 |
| 2 | bge-en-icl_few-shot | 74.63 |
| 3 | BM25 (Reranker: bge-reranker-v2-m3) | 74.2 |
| 4 | BM25 (Reranker: jina-reranker-v2-base-multilingual) | 73.9 |
| 5 | NV-Embed-v2 | 73.45 |
| 6 | bge-en-icl_zero-shot | 73.43 |
| 7 | BM25 (Reranker: bce-reranker-base_v1) | 73.43 |
| 8 | NV-Embed-v1 | 72.08 |
| 9 | bge-multilingual-gemma2 | 71.41 |
| 10 | BM25 (Reranker: gte-multilingual-reranker-base) | 71.36 |
Interactive version: theaggregate.ai/benchmark?slug=air-bench-long-doc · How It Works · Data refreshed daily, snapshot 2026-10-09.