BAGEL - Xeno-canto Call Duration: leaderboard
Metric: Accuracy (%, times 100) on the 649 Xeno-canto questions on call or syllable duration 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 2034. 13 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.4 | 43.9 |
| 2 | Claude Opus 4.6 | 43.5 |
| 3 | Qwen 3 14B (Non-reasoning) | 42.1 |
| 4 | Qwen 3 4B (Non-reasoning) | 34.2 |
| 5 | Mistral 7B Instruct (v0.3) | 32.7 |
| 6 | Llama 3.1 8B Instruct | 32.5 |
| 7 | Qwen 3 8B (Non-reasoning) | 32.2 |
| 8 | gemma-7B | 31.4 |
| 9 | Phi-4 | 30.8 |
| 10 | Qwen 3 32B (Non-reasoning) | 29 |
| 11 | Qwen 3 0.6B (Non-reasoning) | 27.1 |
| 12 | Gemma 3 27B (IT) | 26 |
Interactive version: theaggregate.ai/benchmark?slug=bagel-xeno-canto-call-duration · How It Works · Data refreshed daily, snapshot 2026-10-07.