China's free AI model is giving DeepSeek déjà vu. It works, but takes patience.

A new open-source AI model from China, GLM-5.2, is gaining attention for its capabilities in various tasks, drawing comparisons to the influential DeepSeek model. Developed by Beijing-based Z.ai, this model boasts a remarkable 1 million-token context window, allowing it to process extensive text similar to industry heavyweights like OpenAI's GPT-5.5 and Anthropic's Claude Opus 4.8. However, unlike its paid counterparts, GLM-5.2 is available for free. Testing GLM-5.2 across multiple tasks revealed both its potential and its limitations. The model's speed is notably slower than that of premium AI models, often encountering capacity issues that can lead to frustrating wait times. For example, when tasked with crafting an outreach email for interviews with career coaches, GLM-5.2 delivered a well-structured response that closely resembled a typical email draft, although it required a significant wait. In another task, I sought recommendations for wet cat food suitable for a cat with a sensitive stomach. After overcoming the model's capacity delays, it provided a list of reputable brands and helpful tips for selecting the right food, aligning well with professional veterinary advice. While GLM-5.2 does not yet offer direct shopping links, finding the suggested products online was a quick process. For trip planning, I asked GLM-5.2 to create an itinerary for a weekend getaway from Oakland to Monterey, California. The resulting plan included various activities and destinations, although it initially overlooked the request for budget accommodations. When prompted again, it provided suggestions, but the pricing was not accurate, reflecting a need for improvement in real-time data. The design capabilities of GLM-5.2 also showcased its strengths and weaknesses. Upon requesting an advertisement for a fictional jewelry business, the model initially struggled with capacity issues. After switching to an older version, the output was unexpectedly generated in Chinese, which could confuse users unfamiliar with the language. The final design, while functional, lacked the polish of more expensive models, resembling a basic menu rather than a luxury advertisement. Overall, GLM-5.2 exhibits considerable promise as a free, open-source AI tool, particularly in everyday tasks such as writing, research, and planning. Although it currently falls short of the reliability and responsiveness of its paid counterparts, improvements in performance could position GLM-5.2 as a strong contender for those seeking cost-effective AI solutions.