PhageBench: leaderboard

Interpretation of raw bacteriophage genome sequences on PhageBench (2026), multiple-choice questions built from curated phage databases and answered zero-shot with chain-of-thought prompting, averaged over five tasks: phage contig identification, contamination detection, completeness estimation, lifestyle classification and host prediction. Accuracy in percent; higher is better.

Metric: Accuracy (%) averaged over the five PhageBench tasks (phage contig identification, contamination detection, completeness estimation, lifestyle classification, host prediction), each item a raw-nucleotide multiple-choice question with shuffled options, zero-shot chain-of-thought prompting (reasoning models at reasoning effort medium with at most 2,048 reasoning tokens); higher is better. Source: arxiv.org. Saturation forecast: Around February 2028. 8 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1Gemini 3 FlashMedium56.33#175
2GPT-5.2Medium50.95#185
3Claude Sonnet 4.5Thinking47.88#261
4Qwen 3 MaxThinking45.69#415
5LLaMA 4 (PhageBench checkpoint unspecified)42.6
6GPT-OSS-120B42.27#704
7Qwen3-235B (PhageBench checkpoint unspecified)41.38
8GPT-4o Mini40.15#1667

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

Interactive version: theaggregate.ai/benchmark?slug=phagebench · How It Works · Data refreshed daily, snapshot 2026-11-21.