
Cheaper Chinese AI rivals close the gap and draw U.S. firms toward open stacks
Chinese open-source/open-weight AI models from DeepSeek, Z.ai, and others are narrowing the performance gap with leading U.S. rivals while offering dramatically lower costs, prompting U.S. firms to route more tokens via platforms like OpenRouter and shift workloads away from OpenAI/Anthropic. Lindy moved 100% of Claude traffic to DeepSeek, and GLM 5.2 has seen rapid adoption, highlighting a trend where cheaper Chinese models are becoming viable for many workloads—often 60%–90% cheaper—even as frontier performance remains a few months behind the top U.S. systems, a dynamic occurring amid tighter U.S. AI regulation.













