Carbon-X underwent an embedding warmup training stage using approximately 10 billion tokens with the transformer body frozen to train only new DNA embeddings

Large language models (LLMs) are advanced AI systems trained on massive text datasets to understand, generate, and reason about human language, powering applications such as chatbots, code generation, summarization, and scientific discovery.