The recent release of Zhipu's GLM 5.2 artificial intelligence model has sent shockwaves through the tech community, with its impressive performance and significantly lower cost compared to top US models. On a key benchmark, GLM 5.2 is now within a percentage point of Anthropic's Opus 4.8, achieving this at roughly a fifth of the cost. This has led to a surge in developer interest, with OpenRouter token traffic increasing rapidly. What sets GLM 5.2 apart is its strength in agentic work, such as planning, coding, and testing, making it an attractive solution for enterprises seeking to automate these tasks. The model's open-source nature allows companies to download, fine-tune, and run it on their own servers, putting pressure on closed-source labs. As regulatory oversight increases, with Anthropic and OpenAI recently limiting their models due to government requests, the appeal of an open-source model that cannot be revoked is growing. According to Gabe Pereyra, co-founder of Harvey, 'GLM 5.2 is the first open-source model to be truly competitive with closed-source frontier models.' As companies look to optimize their AI spend, the 'intelligence per dollar' metric is becoming increasingly important, making Zhipu's cost-effective approach a compelling option.

Source: CNBC