MiniMax H1 gross margin 17.9% as M3.1 Flash claims ~3x faster inference
The author links MiniMax's falling gross margin to M3.1 Flash's nearly 3x faster inference, arguing efficiency must pay for price cuts and H2 margin is the number to watch.
Shanghai AI lab MiniMax's first-half gross margin was 17.9%, up from 12.1% a year earlier, but below an implied 34% in the second half of 2025. In August, management said it expects margins to improve in the second half of 2026. MiniMax says M3.1 Flash runs online inference nearly three times faster. My working assumption is that the efficiency behind that speed has to pay for the lower prices. The second-half gross margin is the first number I will check. It is one sign of whether efficiency kept pace with the price cut, though the filing will not separate the two.
This post quotes another post: “MiniMax M3.1 Flash pricing: input and output each over 50% below M3”