LitReviewBench: leaderboard
Metric: Expert-preference rating (Elo scale, Bradley-Terry aggregation of about 3,000 anonymized pairwise battles judged by matched domain experts on D5 overall utility as a researcher's starting point; literature review drafts on AI topics; the human-draft baseline is in the rating pool). Source: arxiv.org. Saturation forecast: Around March 2027. 9 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Sonar Deep Research | 1285.9 |
| 2 | Claude Opus 4.5 | 1135.5 |
| 3 | Qwen 3 235B A22B | 836.2 |
| 4 | Grok 4 | 799.2 |
| 5 | GLM-4.6 | 434.3 |
| 6 | Gemini 2.5 Pro | 400.2 |
Interactive version: theaggregate.ai/benchmark?slug=litreviewbench · How It Works · Data refreshed daily, snapshot 2026-09-26.