China’s GLM‑4.5: Cheaper, leaner, more “agentic” AI
China’s GLM‑4.5: Cheaper, leaner, and more “agentic”
At the 2025 World AI Conference in Shanghai, Chinese AI startup Z.ai (formerly Zhipu) unveiled its new GLM‑4.5 series, a family of open‑source language models designed to undercut rivals on price and efficiency. The headline model runs on just eight Nvidia H20 chips—half the hardware used by competitor DeepSeek’s R1—and Z.ai says it will charge only 11 ¢ per million input tokens and 28 ¢ per million output tokens via its API. In other words, the cost of generating text with GLM‑4.5 is more than ten times lower than DeepSeek’s current rates.
Z.ai also touts an “agent‑native” design, meaning GLM‑4.5 can break complex instructions into subtasks and handle multi‑step reasoning. The company is releasing three sizes: the full GLM‑4.5 with 355 billion parameters, a streamlined “Air” version that activates only a small subset of parameters for efficiency, and a “Flash” model optimized for speed. All are offered under an open, auditable license that allows businesses to deploy them on their own hardware.
At the 2025 World AI Conference in Shanghai, Chinese AI startup Z.ai (formerly Zhipu) unveiled its new GLM‑4.5 series, a family of open‑source language models designed to undercut rivals on price and efficiency. The headline model runs on just eight Nvidia H20 chips—half the hardware used by competitor DeepSeek’s R1—and Z.ai says it will charge only 11 ¢ per million input tokens and 28 ¢ per million output tokens via its API. In other words, the cost of generating text with GLM‑4.5 is more than ten times lower than DeepSeek’s current rates.
Z.ai also touts an “agent‑native” design, meaning GLM‑4.5 can break complex instructions into subtasks and handle multi‑step reasoning. The company is releasing three sizes: the full GLM‑4.5 with 355 billion parameters, a streamlined “Air” version that activates only a small subset of parameters for efficiency, and a “Flash” model optimized for speed. All are offered under an open, auditable license that allows businesses to deploy them on their own hardware.
A booming Chinese AI scene
Z.ai is just one of more than a thousand large language models released by Chinese firms—an industry wave that accounts for roughly 40 % of models worldwide. Tech giants like Alibaba and Tencent have launched their own open‑source models; the former’s Qwen3‑Coder targets code generation, while the latter’s HunyuanWorld focuses on 3‑D content. Startups like Moonshot and DeepSeek are also pushing aggressively into the global market.
This growth has prompted concerns in the West. Some Chinese models have been flagged for routing data back to servers in China, creating privacy risks. Z.ai is on a U.S. trade‑restriction list, and American regulators are tightening export controls on advanced AI chips. At the same time, Washington is crafting an AI Action Plan to boost domestic adoption and guard against losing technological leadership.
Z.ai is just one of more than a thousand large language models released by Chinese firms—an industry wave that accounts for roughly 40 % of models worldwide. Tech giants like Alibaba and Tencent have launched their own open‑source models; the former’s Qwen3‑Coder targets code generation, while the latter’s HunyuanWorld focuses on 3‑D content. Startups like Moonshot and DeepSeek are also pushing aggressively into the global market.
This growth has prompted concerns in the West. Some Chinese models have been flagged for routing data back to servers in China, creating privacy risks. Z.ai is on a U.S. trade‑restriction list, and American regulators are tightening export controls on advanced AI chips. At the same time, Washington is crafting an AI Action Plan to boost domestic adoption and guard against losing technological leadership.
What it means for AI
The release of GLM‑4.5 underscores how fast AI technology is evolving and how intense the competition has become. Cheaper, more efficient models pressure established players in the U.S. and Europe to improve performance and reduce costs. They also encourage wider experimentation with open‑source licensing, which appeals to organizations that want transparency and control over their data.
However, the geopolitical divide between Chinese and Western AI ecosystems is likely to widen as each side pursues its own regulatory and security priorities. That could lead to regional “AI spheres” rather than a single global marketplace. For developers and businesses, the takeaway is clear: new options are emerging at breakneck speed, but choosing the right model will increasingly depend on more than just technical performance—issues like data sovereignty, legal exposure and trust will be just as important.
The release of GLM‑4.5 underscores how fast AI technology is evolving and how intense the competition has become. Cheaper, more efficient models pressure established players in the U.S. and Europe to improve performance and reduce costs. They also encourage wider experimentation with open‑source licensing, which appeals to organizations that want transparency and control over their data.
However, the geopolitical divide between Chinese and Western AI ecosystems is likely to widen as each side pursues its own regulatory and security priorities. That could lead to regional “AI spheres” rather than a single global marketplace. For developers and businesses, the takeaway is clear: new options are emerging at breakneck speed, but choosing the right model will increasingly depend on more than just technical performance—issues like data sovereignty, legal exposure and trust will be just as important.
.png)
Comments
Post a Comment