From：Internet Info Agency 2026-08-26 22:30:00
On August 26, during its interim earnings conference call, MiniMax announced that the parameter scale of its M3Pro model is expected to increase to approximately 3 trillion, enhancing its capabilities in reinforcement learning and long-horizon task training. The company stated it will advance the co-optimization of model intelligence, inference efficiency, and infrastructure capabilities, and accelerate the development and release cycles of its M-series and H-series models. MiniMax is currently advancing the adaptation of its M3 and H3 models to domestically produced chips, with large-scale domestic computing clusters expected to come online soon and gradually take on real production workloads. MiniMax reported financial results for the first half of fiscal year 2026, showing total revenue of $117 million, an increase of 283.1% year-over-year; gross profit of $20.813 million, up 464.8% year-over-year; and a gross margin of 17.9%. Net loss attributable to shareholders was $358 million, representing an 11% year-over-year narrowing of losses. Both basic and diluted earnings per share were -$1.18, and the asset-liability ratio stood at 21.5%.
