A Cheap Chinese AI Model Challenges Anthropic and OpenAI
A cheap Chinese AI model is making headlines by surpassing Anthropic and OpenAI in certain benchmarks. Z.ai’s GLM-5.2 has achieved fourth place on a leading intelligence ranking, demonstrating remarkable capabilities at a fraction of the cost.
Key Findings:
- Performance: GLM-5.2 scores 51 on Artificial Analysis’ Intelligence Index v4.1, outperforming models like MiniMax-M3, DeepSeek V4 Pro, and Kimi K2.6.
- Pricing: It costs as little as $1.40 per million input tokens and $4.40 per million output tokens, making it significantly cheaper than Anthropic’s Claude Opus or OpenAI’s GPT-5.5.
- Technological Features: Trained on domestic Chinese silicon using a cluster of 100,000 Huawei Ascend 910B processors, circumventing US chip restrictions.
Expert Opinion:
David Sacks, former White House AI czar, states that GLM-5.2 is "as good as the currently available models from OpenAI and Anthropic", challenging the US lead in AI development.
Notable Trade-Off:
While scoring high in abstract reasoning puzzles and real-world agentic work, GLM-5.2 uses more output tokens per task than its peers, impacting practical costs.
This development underscores the rapid progress of Chinese AI, offering open-source access to advanced models while potentially disrupting the market dynamics established by Western companies.