Model Comparison

Phi 4 Mini

Microsoft

Phi 4 Mini is a language model developed by Microsoft. It achieves strong performance with an average score of 65.4% across 17 benchmarks. It excels particularly in GSM8k (88.6%), ARC-C (83.7%), BoolQ (81.2%). It's licensed for commercial use, making it suitable for enterprise applications. Released in 2025, it represents Microsoft's latest advancement in AI technology.

Qwen2 72B Instruct

Alibaba

Qwen2 72B Instruct is a language model developed by Alibaba. It achieves strong performance with an average score of 73.6% across 17 benchmarks. It excels particularly in GSM8k (91.1%), CMMLU (90.1%), HellaSwag (87.6%). The model shows particular specialization in code tasks with an average performance of 82.3%. Released in 2024, it represents Alibaba's latest advancement in AI technology.