With this at 1k tps and Kimi 2.6 1k tps by Cerebras, I believe we are entering the next stage of LLMs, where companies will also compete on throughput