Chemical knowledge and reasoning on ChemBench
86.1Overall ScoreS1-NexusAgent
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| S1-NexusAgentModel Category=Agent, Executor=Claude-4.5-Sonnet2026.02 | 86.1 | 91.5 | 84.2 | 96.6 | 95.7 | 91 | 94.4 | 89.6 | 85 | 77.2 | |
| Intern-S1Model Category=Base Model2026.02 | 83.4 | — | — | — | — | — | — | — | — | — | |
| S1-NexusAgentModel Category=Agent, Executor=DeepSeek-V3.22026.02 | 72.1 | 90.1 | 51.7 | 95.3 | 84.8 | 79.8 | 89.8 | 88 | 85.5 | 75.1 | |
| S1-NexusAgentModel Category=Agent, Executor=Qwen3-8B-Sci2026.02 | 67 | 89.8 | 40.4 | 94 | 84.2 | 78.4 | 88.6 | 86.6 | 82.3 | 72.9 | |
| DeepSeekModel Category=Base Model, Backbone=DeepSeek-V3.12026.02 | 66 | 62 | 59 | 92 | 87 | 77 | 85 | 85 | 82 | 48 | |
| ClaudeModel Category=Base Model, Backbone=Claude-3.7-Sonnet2026.02 | 66 | 65 | 60 | 91 | 88 | 76 | 84 | 82 | 80 | 48 | |
| GeminiModel Category=Base Model, Backbone=Gemini-2.5-flash2026.02 | 64 | 62 | 56 | 85 | 84 | 75 | 83 | 75 | 78 | 48 | |
| ReActModel Category=Agent, Backbone=Claude-3.5-Sonnet2026.02 | 63 | 57 | 58 | 83 | 82 | 50 | 82 | 81 | 85 | 44 | |
| S1-NexusAgentModel Category=Agent, Executor=Qwen3-8B2026.02 | 61.5 | 72.8 | 38.9 | 86.6 | 83.2 | 70.3 | 78.9 | 78 | 75 | 69.8 | |
| GPTModel Category=Base Model, Backbone=GPT-4o2026.02 | 61 | 56 | 59 | 81 | 80 | 75 | 76 | 72 | 75 | 44 | |
| ReActModel Category=Agent, Backbone=GPT-4o2026.02 | 51 | 47 | 42 | 76 | 73 | 40 | 72 | 62 | 72 | 37 |