IA Generativa
GLM-5.2 Challenges Claude with 39% F1 Score in IDOR Tests
Dr. Adrian Vale
LLM, AI Agents & AI Infrastructure Specialist
4 min read
Listen to Article
AI automated narration
LLM, AI Agents & AI Infrastructure Specialist
Zhipu AI's open-source model, GLM-5.2, achieved a 39% F1 Score in IDOR vulnerability detection, outperforming Anthropic's Claude at 32%. With a cost of $0.17 per detection, GLM-5.2 delivers a 47% cost reduction compared to proprietary solutions, challenging closed-source models like Claude and GPT while offering significant savings for enterprises.
On June 13, 2026, Zhipu AI launched GLM-5.2, the latest iteration of its open-source AI model, under the MIT license. This move positions GLM-5.2 as a cost-effective solution for enterprises and researchers looking for high-performance AI without the limitations of proprietary platforms. With a design tailored for tasks such as vulnerability detection and long-context processing, GLM-5.2 directly challenges the dominance of closed-weight systems like Anthropic’s Claude and OpenAI’s GPT.
GLM-5.2 has made significant strides in the field of cybersecurity. According to Semgrep, a specialized code analysis platform, GLM-5.2 achieved a 39% F1 Score in detecting IDOR (Insecure Direct Object Reference) vulnerabilities. This score surpasses the 32% F1 Score achieved by Anthropic’s Claude, a leading proprietary model.
This performance improvement is not only a testament to GLM-5.2's technical capabilities but also to its value proposition. At a cost of just $0.17 per vulnerability detected, GLM-5.2 offers a 47% cost savings compared to proprietary solutions, making it an attractive option for organizations aiming to optimize cybersecurity expenditures. Importantly, the evaluations were conducted using standardized datasets and consistent prompts, lending credibility to the benchmarking results.
The introduction of GLM-5.2 under an open-source license is a disruptive force in the AI landscape. Here's why:
The release of GLM-5.2 raises several questions about the trajectory of the AI industry:
The success of GLM-5.2 could mark a turning point, not just in cybersecurity but in how the industry perceives the role of open-source in AI development.
GLM-5.2 is an open-source AI model developed by Zhipu AI, focused on tasks like cybersecurity vulnerability detection and long-context processing.
GLM-5.2 outperformed Claude in IDOR vulnerability detection, achieving a 39% F1 Score compared to Claude's 32%, while also being 47% more cost-effective.
GLM-5.2 detects vulnerabilities at $0.17 per detection, offering a 47% cost savings compared to proprietary models like Claude.
💡 Dica Pro: Utilizing open-source models like GLM-5.2 not only reduces costs but also allows for unique customizations that proprietary solutions can't match. By tuning the model to specific use cases, organizations can achieve superior results tailored to their unique requirements.
