We burned 11.7bn tokens to find the best cyber AI model | GLM5.3 and DeepSeek are now frontier - Aikido Security

Aikido Security burned 11.7 billion tokens benchmarking cyber AI models, finding that GLM 5.3 and DeepSeek have reached frontier-level performance in cybersecurity tasks. The results suggest Chinese open-weight models no

Read full story at deepseek →