MiniMax Plans to Open-Source 2.7 Trillion Parameter Model

Chinese AI startup MiniMax is building a new large language model with 2.7 trillion parameters and plans to release it as open source later this year. That would make it one of the largest open-source models ever, representing an order-of-magnitude leap from existing ones like Llama 3 (405B) or Qwen 2 (72B). For research institutions and well-resourced enterprises, this model offers the chance to deploy an ultra-large LLM locally, especially for tasks requiring deep understanding and generation. However, the 2.7 trillion-parameter model demands extremely high hardware requirements, making it impractical for most developers or small teams to run at affordable cost. It will likely serve more as a research reference or be accessed via API indirectly.