Chinese AI company MiniMax released MiniMax M3.1 Flash on October 19, 2026, featuring 428 billion total parameters, about 23 billion active parameters in a mixture-of-experts architecture, and a 1,000,000-token context window.

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.