Chinese Open-Source AI Models Gain Use on OpenRouter While Trailing Top US Systems on Some Benchmarks

ai released its GLM-5.2 model last month, and the open-source system has since ranked as the fifth most commonly used large language model on OpenRouter. Artificial Analysis placed it as the most intelligent open-source AI available, though the model performed slightly worse than OpenAI’s GPT-5.5 on the firm’s intelligence tests.

ai stated that GLM-5.2 outperforms GPT-5.5 on a benchmark for AI software engineering skills.

DeepSeek’s latest model remains the most used on OpenRouter, with more than twice the traffic of GLM-5.2. Seven of the top 10 models on the platform were built by Chinese companies. A UK software developer reported that a local Z.ai model completed the same task 30 percent slower than an OpenAI cloud model but incurred no cost.

When DeepSeek released its open-source R1 model in January 2025, the system was reported to rival leading US models and was offered free for download. US tech companies lost a trillion dollars in value, and lawmakers proposed banning the model on government devices. The January release produced no comparable market reaction to GLM-5.2.

The US government has maintained and tightened export controls on advanced chips to China in recent years. President Donald Trump has discussed taking a public stake in companies such as OpenAI. New Scientist reported that Chinese firms typically release models for local download at no charge, while most US firms host models in the cloud for a fee and limit access to their internals.

Open-source models from China trail the most expensive US systems on some measures but remain adequate for many users.