Zhipu AI's GLM-5.2 triggers fresh 'DeepSeek moment' in coding AI
Synopsis
Beijing's Zhipu AI released GLM-5.2 on June 13 — the first Chinese open-weight model to crack the global top three on a major AI benchmark — prompting a former Meta and Google DeepMind VP to call it the 'first open model that passes the bar as a daily driver,' echoing the original DeepSeek shock.
Key Takeaways
Zhipu AI (also known as Z.ai ) released open-weight model GLM-5.2 on June 13, 2026 , targeting coding workflows.
GLM-5.2 is the first Chinese AI model to rank in the top three globally on a major AI benchmark, according to available evaluations.
Former Meta Platforms and Google DeepMind vice-president Matt Velloso called it the 'first open model that passes the bar as a daily driver.' The launch came one day after Anthropic pulled Claude Fable 5 from international access to comply with a Washington directive restricting foreign users.
The release is being compared to DeepSeek 's disruptive debut roughly 18 months earlier, which similarly rattled Silicon Valley with its cost-performance ratio.
Competing Chinese models — DeepSeek V4 Pro , MiniMax M3 , and Alibaba 's Qwen3.7-Max — have also made gains but had not previously reached the global top three.
Beijing-based Zhipu AI has ignited fresh debate in global tech circles with the release of GLM-5.2, a cost-effective open-weight coding model that American entrepreneurs and researchers are calling the first Chinese AI to credibly challenge proprietary Western alternatives for day-to-day development workflows. The model, made publicly available on June 13, 2026, is being likened by some observers to the disruptive debut of DeepSeek in early 2025 — a comparison that has reverberated across Silicon Valley and beyond.
A benchmark breakthrough
GLM-5.2 has become the first Chinese model to rank in the top three globally on a major AI benchmark, according to available evaluations. While recent Chinese releases — including DeepSeek V4 Pro, MiniMax M3, and Alibaba's Qwen3.7-Max — have made notable strides, none had previously cracked the upper tier of global rankings. The achievement marks a significant inflection point in the competitive landscape between Chinese and Western AI labs.
Silicon Valley takes notice
Matt Velloso, a former vice-president at both Meta Platforms and Google DeepMind, said on X earlier this month that he had been using GLM-5.2 'all day' and found it to be the 'first open model that passes the bar as a daily driver.' He described the model as 'more to the point, doesn't talk too much, doesn't go in circles trying to explain itself, just does the job' — a direct comparison to OpenAI's proprietary GPT-5.5, released in April. 'Things are not going to be the same,' Velloso wrote.
Why it matters
The timing of GLM-5.2's launch is notable: it arrived one day after leading US lab Anthropic withdrew its most advanced public-facing model, Claude Fable 5, to comply with a Washington directive blocking access for foreign users. That regulatory retreat created an opening in the global open-weight market that Zhipu AI — known internationally as Z.ai — appears to have moved quickly to fill. The episode underscores how US export-control and access policies can inadvertently accelerate the international visibility of Chinese AI alternatives.
The competitive backdrop
Nearly 18 months after DeepSeek rattled Silicon Valley with a powerful yet affordable model, China's AI ecosystem has continued to compress the capability gap with Western labs at a pace that has surprised many industry analysts. Zhipu AI's open-weight approach — making model weights freely available — lowers the barrier for developers globally to adopt and build on its technology, a strategy that mirrors DeepSeek's own playbook and has proven effective at generating rapid grassroots adoption.
What's next
The reception of GLM-5.2 will be closely watched by investors, policymakers, and competing labs as a signal of whether China's AI sector can sustain top-tier benchmark performance while maintaining the cost advantages that defined the original DeepSeek moment. How OpenAI, Anthropic, and Google DeepMind respond — and whether Washington moves to restrict access to open-weight Chinese models — will determine the next chapter of this intensifying rivalry.
Point of View
A dynamic that US policymakers appear not to have fully anticipated. The open-weight strategy is the sharpest competitive weapon here: once model weights are public, no export control can un-release them, which means Washington's toolkit is far less effective against open-source Chinese AI than against chip shipments. The real stress test for GLM-5.2 will come if the US moves to restrict access to the model itself — a step that would force a direct confrontation between open-source norms and national-security imperatives.
NationPress
26 Jun 2026
Frequently Asked Questions
What is Zhipu AI's GLM-5.2 and why is it significant?
GLM-5.2 is an open-weight AI model released by Beijing-based Zhipu AI on June 13, 2026 , designed primarily for coding tasks. It is significant because it became the first Chinese AI model to rank in the top three globally on a major AI benchmark, and prominent US technologists have described it as reliable enough for daily professional coding use.
Why is GLM-5.2 being called a new 'DeepSeek moment'?
The term 'DeepSeek moment' refers to the disruption caused by DeepSeek 's affordable yet powerful model in early 2025 , which shocked Silicon Valley . GLM-5.2 is drawing the same comparison because it combines high benchmark performance with an open-weight, cost-effective approach that challenges proprietary Western models — a pattern observers say could again reshape developer preferences globally.
What did Matt Velloso say about GLM-5.2?
Matt Velloso , a former vice-president at Meta Platforms and Google DeepMind , said on X that he had been using GLM-5.2 'all day' and called it the 'first open model that passes the bar as a daily driver.' He also said the model was 'more to the point, doesn't talk too much, doesn't go in circles trying to explain itself, just does the job,' comparing it favourably to OpenAI 's GPT-5.5 .
How does GLM-5.2's release relate to Anthropic pulling Claude Fable 5?
Anthropic withdrew its most advanced public-facing model, Claude Fable 5 , from foreign users one day before GLM-5.2 launched, complying with a Washington directive. This regulatory retreat created a gap in the global open-weight AI market that Zhipu AI moved quickly to fill, illustrating how US access restrictions can inadvertently boost the visibility of Chinese alternatives.
How does GLM-5.2 compare to other recent Chinese AI models?
Unlike DeepSeek V4 Pro , MiniMax M3 , and Alibaba 's Qwen3.7-Max — which have all made significant benchmark gains — GLM-5.2 is the first Chinese model to reach the top three globally on a major AI benchmark. Its open-weight availability also sets it apart, lowering adoption barriers for developers worldwide and mirroring the strategy that made DeepSeek a global phenomenon.
