SMDD-Bench - 2D Pharmacophore Identification: leaderboard

Success rate (%) on the 25 tasks that ask for the 2D pharmacophore of a ligand series in SMDD-Bench (2026), where LLM agents with a minimal tool harness and a limited budget of Boltz-2 structure and ADMET-AI property oracles must submit molecules that pass hidden evaluators. Higher is better.

Metric: Success rate (%) on the 25 SMDD-Bench 2D pharmacophore identification tasks, minimalist ReAct agent harness with RDKit-style tools, 8 Boltz-2 and 15 ADMET-AI oracle calls per task, temperature 1.0, at most 100 turns; success means the submitted Python pharmacophore function passes recall and specificity checks on hidden active and inactive molecules; higher is better. Source: arxiv.org. Saturation forecast: Around September 2028. 7 models tracked. Updated 14 Nov 2026.

Top models

#ModelSettingScoreOverall rank
1Claude Sonnet 4.628#173
2Gemini 3.1 Pro (Preview)Medium20#99
3MiniMax-M2.716#539
4Kimi K2.5Thinking12#266
5GPT-5.4Medium12#135
6DeepSeek V3.28#419
7Qwen 3.5 397B A17B4#264

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

Interactive version: theaggregate.ai/benchmark?slug=smdd-bench-2d-pharmacophore-identification · How It Works · Data refreshed daily, snapshot 2026-11-21.