Rick W / Monday, March 16, 2026 / Categories: Artificial Intelligence Clarifai Reasoning Engine Achieves 414 Tokens Per Second on Kimi K2.5 Clarifai achieves 414 tokens per second on Kimi K2.5, one of the first providers to reach 400+ TPS on a trillion-parameter reasoning model running on Nvidia B200 GPUs. Previous Article Insights Wherever You Work: Meet the Tableau App for Microsoft 365 Next Article Flash Attention 2: Reducing GPU Memory and Accelerating Transformers Print 0 Tags: ModeModelGPUNVIDIA