MiniMax's fast, low-cost multimodal coding model with a 1M-token context window and tunable thinking depth.