Chinese open models are narrowing the distance to closed systems, while fragmented benchmarking practices obscure the measurement of actual performance differences.
A change to US copyright law could clarify training data usage and ban distillation restrictions, while Chinese manufacturers increasingly pursue open-source releases.
