Property Comparison & Ranking on PolyBench (test)
77MCQ AccuracyGPT-5
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GPT-5category=Closed-source2026.01 | 77 | 90 | |
| GPT-4ocategory=Closed-source2026.01 | 69 | 86 | |
| Sonnetcategory=Closed-source2026.01 | 66 | 85 | |
| PolyLM (Phi-4-14B)category=PolyBench, CoT=true2026.01 | 65 | 71 | |
| PolyLM (Qwen-2.5-14B)category=PolyBench, CoT=true2026.01 | 63 | 73 | |
| Phi-4-14Bcategory=Off-the-shelf2026.01 | 61 | 70 | |
| PolyLM (ChemLLM-7B)category=PolyBench, CoT=false2026.01 | 59 | 69 | |
| PolyLM (Phi-4-14B)category=PolyBench, CoT=false2026.01 | 57 | 85 | |
| Qwen-2.5-14Bcategory=Off-the-shelf2026.01 | 55 | 69 | |
| 4o-Minicategory=Closed-source2026.01 | 55 | 68 | |
| PolyLM (ChemLLM-7B)category=PolyBench, CoT=true2026.01 | 54 | 50 | |
| PolyLM (Qwen-2.5-14B)category=PolyBench, CoT=false2026.01 | 50 | 87 | |
| PolyLM (Qwen-2.5-7B)category=PolyBench, CoT=false2026.01 | 49 | 60 | |
| PolyLM (Qwen-2.5-7B)category=PolyBench, CoT=true2026.01 | 48 | 57 | |
| ether0-24Bcategory=Domain aligned2026.01 | 46 | 51 | |
| Qwen-2.5-7Bcategory=Off-the-shelf2026.01 | 43 | 58 | |
| ChemLLM-7Bcategory=Domain aligned2026.01 | 42 | 47 | |
| Llamole-8Bcategory=Domain aligned2026.01 | 39 | 56 | |
| LlaSMol-8Bcategory=Domain aligned2026.01 | 20 | 49 |