Table 4 Base LLMs adopted in this study... | Emerald Publishing

Table 4

Base LLMs adopted in this study

	Qwen2.5–72B	Qwen2.5-MAX	Qwen3-235B-A22 B	DeepSeek-V3	DeepSeek-R1	GPT-4o
# Architecture	Dense	MoE	MoE	MoE	Dense	Dense, Multimodal
# Total Params	72B	325 B	235B	671 B	260B	∼1.8 T (estimated)
# Activated Params	72B	22 B	22B	37 B	260B	∼1.8 T (estimated)
# Availability	Open Source	Open Source	Open Source	Open Source	Open Source	Closed Source
# Model Category	General	General	Reasoning	General	Reasoning	General

	Qwen2.5–72B	Qwen2.5-MAX	Qwen3-235B-A22 B	DeepSeek-V3	DeepSeek-R1	GPT-4o
# Architecture	Dense	MoE	MoE	MoE	Dense	Dense, Multimodal
# TotalParams	72B	325 B	235B	671 B	260B	∼1.8 T (estimated)
# ActivatedParams	72B	22 B	22B	37 B	260B	∼1.8 T (estimated)
# Availability	Open Source	Open Source	Open Source	Open Source	Open Source	Closed Source
# ModelCategory	General	General	Reasoning	General	Reasoning	General

Note(s): MoE is short for mixture-of-expert

Source(s): Authors’ own work