MiniMax founder, chairman and CEO Yan Junjie, in his first English-language podcast interview with British investment trust Scottish Mortgage's "Invest in Progress" podcast, said the company will keep opening up model weights until artificial general intelligence (AGI) is achieved and that profitability will come not from raising model prices but from driving down inference costs through technological innovation. He described MiniMax's mission as minimizing inference costs while putting frontier intelligence in as many hands as possible, noting the name carries the dual meaning of "minimum cost" and "maximum intelligence." On resources, he said compute constraints have existed since the company's first day, with top US labs able to run large numbers of experiments in parallel backed by abundant capital and chips while Chinese companies must allocate every unit of compute more carefully, shaping its technology road map and organizational culture. Yan said open source is a catalyst for commercialization and further innovation, citing multimodal model H3: within two weeks of release, external developers built roughly 300 open models on top of it, including US company fal.ai, which continued training on H3 to launch H3 Max, boosting generation speed to five seconds of video in three seconds. He added open weights let enterprise and government customers deploy models on their own hardware without data leaving their premises.