xbench-ScienceQA - BoN (N=5): leaderboard

The best-of-five (BoN, N=5) score xbench reports beside the five-run average on its graduate-level science questions (search off, gemini-3.5-flash judge), so it rewards a model that gets a question right in any of five tries; the same two ambiguous rows (GPT-6 Astra and Deepseek-V4-Flash) are held.

Metric: Best-of-5 score (%). Source: xbench.org. Saturation forecast: Not forecast. 28 models tracked. Updated 19 Nov 2026.

Top models

#ModelSettingScoreOverall rank
1Kimi K386#61
2Gemini 3 Pro (Preview)85#120
3Claude Opus 4.884#64
4Claude Opus 5.583#9
5Claude Fable 583#32
6Qwen 3.8 Max82#66
7DeepSeek V4.1 FlashMax82#69
8Gemini 3.7 Flash81#42
9DeepSeek V4 ProMax81#198
10Grok 4.581#90
11Qwen 3.8 Max (Preview)81#80
12GLM-5.380#73
13Hy380#211
14GPT-5.5 ProHigh80#40
15GPT-5.479#138

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=xbench-scienceqa-bon-n-5 · How It Works · Data refreshed daily, snapshot 2026-11-20.