BAGEL: leaderboard

BAGEL (2026) tests animal knowledge in closed-book multiple choice: 11,852 questions built from Wikipedia species articles, GloBI ecological interaction records, bioRxiv preprints and Xeno-canto recordings. Accuracy over all questions in percent, higher is better.

Metric: Accuracy (%, times 100) on all 11,852 questions (item-weighted over the four source domains) of BAGEL, closed-book four-option single-answer multiple choice (the source passage is withheld), one unified prompt, greedy decoding with seed 0, a single run per model; questions written by GPT-4o-mini from Wikipedia species articles, GloBI interaction records, bioRxiv preprints and Xeno-canto recordings, options shuffled; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 13 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1GPT-5.476.01#138
2Claude Opus 4.675.87#114
3Gemma 3 27B (IT)67.89#1341
4Llama 3.1 8B Instruct65.22#4655
5Phi-465.11#1951
6Qwen 3 14BNon-reasoning64.99#1284
7Qwen 3 32BNon-reasoning64.41#983
8Qwen 3 8BNon-reasoning62.53#1890
9Qwen 3 4BNon-reasoning57.53#2379
10Mistral 7B Instruct (v0.3)55.32#6759
11gemma-7B50.46#7115
12SmolLM2-360M24.76#9572
13Qwen 3 0.6BNon-reasoning24.64#8357

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

Interactive version: theaggregate.ai/benchmark?slug=bagel · How It Works · Data refreshed daily, snapshot 2026-11-22.