LISABench 2026's first test of the year: Seven top AI models compete for the title of Web3 vulnerability detection king.

On January 5th, with the arrival of 2026, the AI ​​security field ushered in its first major benchmark test. LISABench announced today the launch of its Q1 2026 evaluation, aiming to determine the top-performing cutting-edge AI model in Web3 smart contract vulnerability detection through practical testing. In addition, LISABench is simultaneously launching a community voting prediction reward program.

This evaluation boasts a prestigious lineup, bringing together seven of the world's top cutting-edge models: KIMI K2 (Moonshot AI), DeepSeek V3.2, QWen 3 30b-a3b (Alibaba Cloud), GLM 4.6 (Zhipu AI), GPT-5.2 (OpenAI), Gemini-3-pro-preview (Google), and Claude 4.5 Sonnet (Anthropic).

Currently, voting for the Q1 winner is open. Meanwhile, the LISABench benchmark codebase has been open-sourced on GitHub for developers to review and reproduce.