MiniMax releases M3.1 Flash model for agents with 70% lower costs
The model achieves top-tier performance in coding, cybersecurity, and computer use benchmarks while delivering inference speeds over 200 tokens per second through enhanced sparsity optimization.