List your company
ResearchFormerly WizWand

Human-level Standardized Exam Evaluation on AGIEval

51.05ScoreCPT
15.596424.800734.00543.2093Sep 10, 2024Dec 13, 2024Mar 18, 2025Jun 20, 2025Sep 23, 2025Dec 26, 2025Mar 31, 2026
Updated 6mo ago

Evaluation results

MethodLinks
CPTSize=70B, Evaluation Mode=zero-shot2024.0951.05
Llama-3Size=70B, Evaluation Mode=zero-shot2024.0950.83
Qwen-3Model Type=Open-weight, Number of Parameters=4B2026.0345.87
Qwen-3.5Model Type=Open-weight, Number of Parameters=4B2026.0344.89
Llama-3Size=8B, Evaluation Mode=zero-shot2024.0937.55
CPTSize=8B, Evaluation Mode=zero-shot2024.0937.44
Qwen-2.5Model Type=Open-weight, Number of Parameters=3B2026.0337.15
OLMO-3Model Type=Fully-open, Number of Parameters=7B2026.0333.75
OLMO-2Model Type=Fully-open, Number of Parameters=7B2026.0331.78
YulanModel Type=Fully-open, Number of Parameters=2.4B2026.0328.22
daVinciModel Type=Fully-open, Number of Parameters=3B2026.0326.77
LLaMa-3.2Model Type=Open-weight, Number of Parameters=3B2026.0322.72
Nemotron-CC_ASIPre-training Curation Strategy=AI Discovered, Model Parameters=3B, Training Token Budget=500B tokens2026.0318.3
Nemotron-CC_ASI+Pre-training Curation Strategy=AI Discovered, Model Parameters=3B, Training Token Budget=500B tokens2026.0318.21
Nemotron-CCPre-training Curation Strategy=AI Discovered, Model Parameters=3B, Training Token Budget=500B tokens2026.0318.15
DCLMPre-training Curation Strategy=Human Discovered, Model Parameters=3B, Training Token Budget=500B tokens2026.0317.9
Ultra-FinewebPre-training Curation Strategy=Human Discovered, Model Parameters=3B, Training Token Budget=500B tokens2026.0317.72
Fineweb-EduPre-training Curation Strategy=Human Discovered, Model Parameters=3B, Training Token Budget=500B tokens2026.0316.96