China’s Zhipu is booming with Anthropic and OpenAI held back
ZhipuThe GLM 5.2 AI model was released last week with much Silicon Valley excitement following the launch of DeepSeek last year.
The new open source model of the Chinese artificial intelligence startup has surpassed all other open versions and is now percentage point At roughly one-fifth the cost of Anthropic’s Opus 4.8, according to a closely watched agency benchmark.
Developers come together as OpenRouter token traffic increases faster More than it was after DeepSeek’s V4 launch in April. Unlike DeepSeek, which the market eventually dismissed as a one-time chatbot shocker, GLM 5.2 is strong at the kind of agency work that businesses are racing to automate—planning, coding, testing, and looping.
Impacted by unexpectedly high AI spending on tokens, the measure of data processed and produced by AI models, companies are now asking how they can get the best value for their money.
The most important metric becomes intelligence per dollar, which makes Zhipu’s cheap but good enough model an attractive answer.
“I am constantly amazed at how quickly open source is catching up.” Gabe PereyraHarvey’s co-founder told CNBC. “GLM 5.2, you see the first model that is truly competitive with some of these closed-source frontier models.”
But the real story beneath the price tag is the open source AI moment.
The fact that GLM 5.2 can be downloaded for free, tweaked, and run on a business’s own servers puts price pressure on frontier labs while also making access look shaky. Anthropic was forced to withdraw its Fable Mythos class model at the behest of the Trump administration, and OpenAI announced on Friday that it was limiting GPT 5.6 models due to government request.
Federal oversight has made a model that no one can cancel look increasingly like a safe bet.
