Molecular Property Filtering on MolBench-MS
98AccuracyClaude Code (vanilla)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Claude Code (vanilla)Model Category=Agent, LLM Backend=Claude Sonnet 4.62026.04 | 98 | 99.3 | |
| MolClaw (Claude Code)Model Category=Agent, LLM Backend=Claude Sonnet 4.62026.04 | 98 | 99.3 | |
| OpenClaw (vanilla)Model Category=Agent, LLM Backend=Claude Sonnet 4.62026.04 | 96 | 99.1 | |
| MolClaw (OpenClaw)Model Category=Agent, LLM Backend=Claude Sonnet 4.62026.04 | 96 | 98.9 | |
| BiomniModel Category=Agent2026.04 | 74 | 82.7 | |
| Kimi 2.5Model Category=LLM2026.04 | 48 | 68.7 | |
| Qwen 3.5Model Category=LLM2026.04 | 48 | 71.9 | |
| GLM 5Model Category=LLM2026.04 | 36 | 63.3 | |
| Claude Sonnet 4.6Model Category=LLM2026.04 | 32 | 56.5 | |
| Deepseek V3.2Model Category=LLM2026.04 | 20 | 34.9 | |
| GPT 5.2Model Category=LLM2026.04 | 8 | 11.2 | |
| Gemini 3Model Category=LLM2026.04 | 8 | 21.4 | |
| Minimax 2.5Model Category=LLM2026.04 | 8 | 15.5 |