Back to today’s briefing

Story perspectives

Microsoft Launches BitNet: 2B4T Model Revolutionizes Efficiency

4/20/2025

33 2

Summary

Story summary

Microsoft has unveiled BitNet b1.58 2B4T, a groundbreaking 1-bit large language model boasting 2 billion parameters. This innovative model, optimized for CPU efficiency—including Apple’s M2 chip—was trained on a staggering 4 trillion tokens, surpassing traditional models in speed and memory efficiency. Excitingly, Microsoft plans to expand its capabilities with even larger models in the future.

1 of 1

Story summary

Microsoft has unveiled BitNet b1.58 2B4T, a groundbreaking 1-bit large language model boasting 2 billion parameters. This innovative model, optimized for CPU efficiency—including Apple’s M2 chip—was trained on a staggering 4 trillion tokens, surpassing traditional models in speed and memory efficiency. Excitingly, Microsoft plans to expand its capabilities with even larger models in the future.