Herman Jin · 2026-10-16 · 帖子

I don’t understand why China would want to shoot itself in the foot by restricting the next generation its open-weight models Market is already treating the very negative impact as a done deal and won’t care about later flexibilities — after all, CSPs can only compete with the two labs by relying on these models at the moment; On the other hand, many of DeepSeek’s own optimizations also heavily depend on contributions from the open community However, for CSPs that have already seen the benefit of these models for their enterprise AI, the only real path forward is to TRAIN their own models rather than keep depending on others