The AI community is abuzz with the release of Grok-1, a new Mixture-of-Experts (MoE) model developed by xAI, under the leadership of Elon Musk, and in collaboration with OpenAI and Google. Grok-1 boasts an impressive 314 billion parameters, making it significantly larger than its predecessor, GPT-3.5, and marks a notable advancement in AI accessibility. Despite its size, there are concerns regarding its performance, with some noting it does not significantly outperform smaller models. The model, which is based on MoE architecture with 8 experts and 2 active at any time, is licensed under Apache 2.0 and was trained using JAX and Rust. It finished training in October 2023 and achieved a 73% score on the MMLU benchmark. Its release has been met with mixed reactions, with some praising the open-source approach, including the availability of its weights on Hugging Face, while others critique it for not providing enough information to fully reproduce it or for its performance not meeting expectations.