[標題]最新消息

Test Results of the October 2025 Open-Source Models

Sorting by Developer Region: Light orange represents European models; light blue represents U.S. models; light green represents domestic models; and light purple represents Chinese models.

Percentage Labeling: Values above 50% are highlighted in green, while values below 50% are highlighted in pink.

  • Language Model Benchmarks / Small Models (13B and below)


 
  • Language Model Benchmarks / Large Models (above 13B)