Three Models, One Codebase: Mistral Large 4 Inside Argus Systems
We gave PHP's source code to Argus Systems three times and swapped the model each run: Mistral Large 4, Z.ai GLM 5.3 and DeepSeek V4 Pro. Mistral Large 4 covered the most code, ran the deep stages most often and returned the most Critical and High findings. It also grades hot and writes long, and we want to talk about both.