SIN-Bench: leaderboard

Metric: Average Overall Score (%; mean of the SIN-Find, SIN-Verify, SIN-QA and SIN-Summary task scores). Source: arxiv.org. Saturation forecast: Around July 2028. 8 models tracked.

Top models

#ModelScore
1Gemini 3 Pro (Preview)56.6
2Claude Sonnet 4.5 (Thinking)54.9
3GPT-554.4
4Gemini 2.5 Pro51
5Grok 449.5
6Qwen 3 VL 8B (Thinking)45.2
7Qwen 3 VL 30B A3B (Thinking)44.8

Interactive version: theaggregate.ai/benchmark?slug=sin-bench · How It Works · Data refreshed daily, snapshot 2026-09-25.