Zhipu AI tops HK$1 trillion after open-source model rivals US leaders

Zhipu AI’s GLM-5.2 Challenges Top US Models at a Fraction of the Cost

Zhipu AI’s market capitalization surpassed HK$1 trillion on Monday after its open-weight GLM-5.2 model emerged as the highest-performing open-source large language model in the world, scoring within a single point of Anthropic’s flagship on key coding benchmarks while priced 72 to 82 percent lower.

A New Open-Source Frontier

Released on June 16, GLM-5.2 is a 753-billion-parameter Mixture-of-Experts model that activates roughly 39 billion parameters per token, supports a one-million-token context window, and ships under an MIT license. On FrontierSWE, a benchmark measuring multi-hour autonomous software engineering projects, GLM-5.2 scored 74.4 — trailing Anthropic’s Opus 4.8 at 75.1 by less than a point while surpassing OpenAI’s GPT-5.5 at 72.6. It is the first open-weight model to compete directly with closed-source leaders on long-horizon coding tasks, according to independent evaluator Artificial Analysis.futunn

The model’s API pricing starts at $1.40 per million input tokens and $4.40 per million output tokens, compared with $5.00 and $25.00 respectively for Opus 4.8 — a discount of roughly 72 percent on inputs and 82 percent on outputs. VentureBeat reported that GLM-5.2 beats GPT-5.5 on multiple long-horizon coding benchmarks “for one-sixth the cost.”llm-stats

Trillion-Dollar Milestone

Shares of Zhipu AI, listed in Hong Kong as Knowledge Atlas Technology, surged as much as 42 percent on Monday, pushing the company’s market capitalization past HK$1 trillion (US$128 billion) before settling at a 15.1 percent gain. The stock has risen more than 1,700 percent since its January debut, according to the South China Morning Post.scmp

JPMorgan raised its price target on Zhipu to HK$1,400 earlier this month, citing the firm’s model visibility and pricing power. In a subsequent note, the bank characterized capital flows into Chinese AI stocks as a “substitution trade” concentrated on Chinese assets rather than a panic-driven sell-off of US equities — a distinction from the DeepSeek shock of early 2025.straitstimes

Regulatory Backdrop Amplifies the Narrative

The timing has amplified comparisons to a “DeepSeek 2.0 moment.” The US Department of Commerce recently compelled Anthropic to suspend global access to its Fable 5 and Mythos 5 models under export control regulations, crystallizing what analysts describe as a “China opening up versus US restrictions” dynamic. JPMorgan noted that enterprises dependent on closed-source frontier models now face business continuity risks, increasing demand for open-weight alternatives that can be deployed locally.futunn

The convergence of a near-frontier open-source model priced at a fraction of its proprietary rivals — arriving just as Washington restricts access to those rivals — has left trading desks reassessing whether the competitive structure of the AI supply chain is undergoing a lasting shift.futunn