How: six GPUs, each running the same AI, writing 1,536 parts of the book at the same time.
No single one is fast. Together they hit 20,000 tokens a second.
The video, the whole book and how we measured it: puretensor.ai/blog/twenty-th…
We hit 20,000 tokens a second and wrote an entire 1,477-page book in 44 seconds.
One prompt. One open model, Qwen3.8-27B by @Alibaba_Qwen. Six @nvidia GPUs in our own lab.
Watch it happen. 👇
#AI#LocalLLM#Qwen#OpenSource