Chinese artificial-intelligence startups are gaining traction in the U.S. as cheaper, increasingly capable models lure developers and cost-conscious companies. Moonshot’s Kimi K3 and Zhipu’s GLM-5.2 have won converts—including Mozilla’s CTO—on speed and price, with OpenRouter data and app-download figures showing surging interest. The economics of “agentic” AI—workflows that generate many tokens—are amplifying the savings versus higher-priced rivals from OpenAI and Anthropic. Washington has tightened export controls and raised concerns about model “distillation,” even as U.S. restrictions have at times created openings for Chinese entrants. Analysts note Chinese models still trail the top U.S. systems on full-range capability and face sustainability questions amid heavy losses, but their open-source tilt and rapid iteration position them for broader global adoption absent an outright ban.
Related articles:
Open-source artificial intelligence
Large language model
OpenRouter model directory and popularity rankings
Qwen: Alibaba’s open-source large language models
DeepSeek AI GitHub organization
