Corridor's Argos 1 reportedly ranks first among 14 models tested on its security-context benchmark

LangChain says Corridor's fine-tuned 9B Argos model scored highest among the 14 models Corridor tested on its security-context benchmark. It says Corridor built the training data with smithtune's trace-based workflow. A separate post about the Argos 1 release claims Corridor fine-tuned Qwen3.5-9B and that the resulting model beat GPT-6, Opus 5.5 and Gemini 3.6 Flash on quality and response time when selecting relevant security guidance for coding agents.