Grok-1 is xAI’s first large language model, built by Elon Musk’s AI lab and originally offered to X Premium+ subscribers starting in November 2023. It is a 314 billion parameter mixture-of-experts model with 8 experts and top-2 routing, meaning only about a quarter of its weights activate for any given token, and it uses an 8,192-token context window. xAI open-sourced the base model weights and architecture on March 17, 2024 under the Apache 2.0 license, making Grok-1 the largest open-weight MoE model publicly available at the time. The release included only the raw pretrained weights, not an instruction-tuned or chat-ready version, so the model was mainly picked up by researchers and infrastructure teams rather than end users. xAI trained Grok-1 on a custom JAX and Rust stack instead of PyTorch, and pretraining had concluded by October 2023, before Grok-2 superseded it later in 2024.