Inferact reports Kimi K3 at 391 tokens/s per user on vLLM TPU v7

Inferact says Kimi K3 running on vLLM with TPU v7 reaches 391 tokens/s per user on SemiAnalysis AgentX. The result uses Inferact's TPU megakernels in vLLM TorchTPU on long-context agentic workloads.

Inferact says the results are publicly viewable on InferenceX and that the megakernels are open source.