Beijing-based AI company Zhipu AI, operating under its Z.ai brand, has released GLM-5.2, its most capable open-source model to date, directly challenging the closed proprietary AI models from OpenAI and Anthropic. The announcement sent Zhipu AI’s shares surging 33% as Wall Street recognized the strategic opportunity presented by the launch.
GLM-5.2 features a one-million-token context window — one of the largest available in any open-source model — along with 128,000 maximum output tokens and two distinct thinking-effort presets optimized specifically for coding tasks. The model is immediately available to subscribers of Z.ai’s GLM Coding Plan, with full API access and a complete open-source release under the permissive MIT license scheduled for later this week.
What makes the GLM-5.2 launch particularly significant is its pricing structure. While Zhipu has not published formal benchmark scores — an unusual but deliberate choice — early community testing suggests performance comparable to Claude Opus 4.7 on coding tasks. Yet the model runs at approximately $1 per million input tokens and $3.20 per million output tokens, compared to $1.75 and $14 respectively for OpenAI’s GPT-5.2. This represents a cost reduction of five to seven times on output-heavy workloads, making frontier-level AI dramatically more accessible.
GLM-5.2 was swiftly added to the Arena AI Agent leaderboard following its release, joining a growing ecosystem of competitive Chinese open-source models that includes DeepSeek and Minimax. JPMorgan analysts subsequently picked Zhipu over MiniMax as their preferred exposure to the Chinese AI open-source wave, citing GLM-5.2’s favorable pricing power and strong adoption potential.
Zhipu AI founder Jie Tang framed the release as a statement about the future of AI development, writing that “the path to AGI should not be surrounded by high walls.” The company’s philosophy aligns with the broader open-source AI movement that argues transparency and accessibility are essential for responsible and equitable development of transformative AI technology.
For enterprises and developers worldwide, the arrival of GLM-5.2 with MIT licensing means more choice, lower costs, and greater flexibility in building AI-powered applications. The ability to run a frontier-class model under an open license, without per-query costs tied to proprietary APIs, opens new possibilities for companies seeking to integrate advanced AI into their products while maintaining data privacy and cost predictability.
