Phi-4: Microsoft's Small AI Model Beats the Giants at Math

This was originally published on post
Its performance on many widely used reasoning-related benchmarks meets or exceeds that of Llama-3.1-405B