According to a report from The Information citing sources familiar with the matter, MiniMax is developing a 2.7 trillion parameter large language model that could be released as soon as the third quarter, and it plans to open source it. The model is internally codenamed M3 Pro, though the final release name hasn't been decided yet.
If the company follows through on releasing the weights, the model would be the largest open source LLM by parameter count. MiniMax's current flagship model, M3, has 428 billion parameters, meaning the new model would be about 6.3 times larger. It would also surpass several existing trillion-parameter open-weight models.
Larger parameter counts generally help with complex reasoning and multi-step instructions, but the final performance will depend on training quality, active parameters, inference cost, and benchmark results. MiniMax will need this model to prove it still belongs in the top tier of Chinese open source models.