Omni-I2C (SMILES) - Tanimoto Similarity: leaderboard

Metric: Tanimoto similarity (0-100) between Morgan fingerprints of the generated and reference SMILES over the 100 chemical-structure images; greedy decoding; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 13 models tracked.

Top models

#ModelScoreOverall rank
1Qwen 3 VL 235B A22B Instruct66.9#264
2Claude Sonnet 4.559.4#138
3Qwen 3 VL 8B Instruct53.4#401
4Qwen 3 VL 32B Instruct51.6#276
5GPT-5.149#131
6Gemini 3 Pro48#77
7Qwen 2.5 VL 72B Instruct29.8#364
8Gemma 3 27B (IT)23.1#509
9Qwen 2.5 VL 7B Instruct15.3#643

No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.

Interactive version: theaggregate.ai/benchmark?slug=omni-i2c-smiles-tanimoto-similarity · How It Works · Data refreshed daily, snapshot 2026-10-11.