Encyclo-K - Science: leaderboard

Encyclo-K's multi-statement selection questions restricted to Science: accuracy over the 1,242 Science questions of the 5,038-question standard set.

Metric: Accuracy on the 1,242 Science questions (%). Source: encyclo-k.github.io. Saturation forecast: Around February 2027. 37 models tracked. Updated 17 Nov 2026.

Top models

#ModelSettingScoreOverall rank
1Qwen 3 235B A22B 2507 (Thinking)55.88#469
2Qwen 3 235B A22B 2507 Instruct53.3#618
3DeepSeek V3.1 TerminusThinking51.61#440
4Kimi K2 090547.99#564
5Qwen 3 30B A3B 2507 (Thinking)43.56#765
6QwQ-32B41.79#908
7Qwen 3 235B A22BThinking41.38#631
8Qwen 3 30B A3B 2507 Instruct38.97#1091
9Qwen 3 32BThinking37.04#965
10Qwen 3 14BReasoning32.61#1264
11Qwen 3 30B A3BThinking32.13#1155
12Qwen 3 4B 2507 Instruct29.79#2215
13DeepSeek V3 (0324)28.58#722
14Qwen 3 8BThinking28.26#1868
15Qwen 3 4BReasoning26.25#2337

No result here: #1 Claude Fable 5.2, #2 GPT-6.2 Astra, #3 Gemini 4 Carbon.

Interactive version: theaggregate.ai/benchmark?slug=encyclo-k-science · How It Works · Data refreshed daily, snapshot 2026-11-17.