BioTool - NCBI: leaderboard
The 476 NCBI queries (E-utilities and BLAST) of the BioTool test split (Gao et al., UC San Diego, 2026): the model writes the API call, which is executed and scored by MedCPT embedding similarity between its returned observation and the reference one (0 for a failed call). Mean in percent, higher is better.
Metric: BioTool score (%; mean MedCPT embedding similarity between the observation returned by the model's API call and the reference observation, 0 for a failed call; the 476 NCBI queries of the test split). Source: arxiv.org. Saturation forecast: Around August 2028. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.5 | 79.8 | #244 |
| 2 | Gemini 3 Pro | 72.5 | #135 |
| 3 | Qwen 3 4B 2507 Instruct | 53.4 | #2174 |
| 4 | Qwen 2.5 7B Instruct | 48 | #2583 |
| 5 | Llama 3.1 8B Instruct | 28.4 | #4277 |
| 6 | Qwen 3 8B | 26.2 | #1838 |
| 7 | GPT-5.1 | 15.5 | #225 |
| 8 | GPT-5.1 Codex | 14.8 | #279 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=biotool-ncbi · How It Works · Data refreshed daily, snapshot 2026-11-12.