April 12, 2026

XAI’s Grok 4.2 boasts 0.5T parameters, Colossus 2 trains 7 models up to 10T

xAI has announced that its current Grok 4.2 model has 0.5 trillion parameters, which is noted to be half the size of its competitor Sonnet and one-tenth the size of Opus, yet it still performs strongly relative to these larger models. This development is part of xAI’s strategy to use the Colossus 2 supercluster, which enables the training of multiple models ranging from 1 trillion to 10 trillion parameters in parallel, positioning the company competitively against leading AI labs.

xAI: xAI is an AI company founded by Elon Musk dedicated to advancing scientific discovery through the development of the Grok series of large language models. It operates Colossus, the world’s largest AI training supercomputer cluster in Memphis, enabling rapid iteration on advanced AI systems. In the news, xAI is advancing its current Grok 4.2 model while using Colossus 2 to train multiple frontier models simultaneously, including language and image generation variants.
Elon Musk: Elon Musk is the founder and leader of xAI, alongside heading Tesla and SpaceX. He frequently provides updates on xAI’s AI developments via X posts. In this instance, Musk confirmed details about the Grok 4.2 model’s size and its competitive performance against Anthropic’s Sonnet and Opus models.

`json
{
“Model Performance”: “Grok 4.2 exhibits impressive capabilities for its size, holding its own against larger models such as Sonnet and Opus.”,
“Development Strategy”: “xAI leverages the Colossus platform to accelerate parallel model training, establishing a competitive stance among leading AI labs.”,
“Training Infrastructure”: “The Colossus 2 infrastructure enables xAI to conduct simultaneous training of advanced AI models, focusing on compute-intensive experimentation with diverse architectures.”
}
`

Sources