Woofun AI reports that Sequoia Capital partners Dean Meyer and Konstantine Buhler observe US leadership in proprietary models while Western firms increasingly utilize Chinese open-weight models like Qwen, Kimi, GLM, and DeepSeek for product bases and synthetic data. ATOM data indicates Qwen’s share in monthly new fine-tuning rose from 1% in January 2024 to 69% in February 2026, with Thinking Machines’ Inkling also employing Kimi K2.5 for early-stage synthetic data generation.

The analysis highlights "distillation" constraints, noting OpenAI and Anthropic restrict output usage for competing training, whereas US entities can legally leverage Chinese open models. Meyer and Buhler propose US labs sell controlled, auditable training rights to retain influence. Although China has not finalized policies restricting overseas access to advanced models, the authors warn Western firms may lag if denied future versions, potentially ceding the open-model foundation to China.