You may not be able to access Anthropic’s best AI model anymore, but open-source alternatives from China appear to be catching up.
What happened: GLM 5.2, a new open-source AI model released earlier this month by the Chinese startup Z.ai, has quickly climbed leaderboards that chart how often different LLMs are being used. It’s also performed well on benchmark rankings that attempt to compare the capabilities of different models.
Early adopters say that GLM 5.2 is particularly good at coding and using agents, two of the most important capabilities for businesses using AI.
What they’re saying: One respected industry observer called GLM 5.2 a “step change” in capabilities and the first “open weight model that feels right in coding harnesses as a general agent.”
The CEO of Vercel, a cloud computing company geared towards developers using AI, posted on X that he was “almost shocked at how good GLM-5.2 by @zai_org is at coding.”
Why it matters: GLM 5.2 may be the first open-source LLM that’s a realistic option for businesses to use. It may not be as powerful as the best closed models offered by Anthropic, OpenAI, and Google DeepMind, but it may be good enough — and it’s much cheaper.
Open-source LLMs can be downloaded for free and run on a user’s own hardware — in theory, they could be used for just the cost of the energy and computing power needed to run them.
Our take: As open-source models improve, it’s very likely the U.S. government will attempt to limit their usage, particularly for those made in China. It’s difficult to imagine a scenario where the White House blocks the latest releases from American AI labs while permitting unrestricted access to Chinese alternatives.
