Scaling up instruction data from the web for to build better LLMs
The strongest open-source LLM for reasoning