Artificial Analysis found MiniMax M3.1 Flash generated approximately 189 million output tokens during its Intelligence Index evaluation, compared to roughly 324 million for GLM-5.3 Flash and 334 million for Gemini 3.8 Flash.