Re:DeepSeek just open-sourced its inference engine
by carstonh on Tuesday October 20, 2026 @02:42PM
(#50170781)
155T tokens/day per cluster? OpenRouter reports that last week total usage on their platform was 164T tokens. Surely all that demand couldn't be satisfied with one cluster? Are these different units of token, or maybe something got lost in translation?
[ Parent
| on HN ]
Re:DeepSeek just open-sourced its inference engine
by yfontana on Tuesday October 20, 2026 @04:18PM
(#50172066)
From the Github benchmark numbers, one GPU can do about 26k tokens/s when serving 256 users, which is about 2.25B tokens/day. A cluster would need to be close to 70k GPUs to serve 155T tokens/day (which seems doubtful).
Re:DeepSeek just open-sourced its inference engine
by brunoff on Tuesday October 20, 2026 @02:47PM
(#50170846)
Exactly... I was going to post exactly this and you called it two minutes before me. On openrouter, it is shown that 600T+ tokens are provided per month, so 20T+ per day. Openrouter is not all the world but it seems that something is not right with this 155T figure... Unless everyone is idle everywhere, waiting for demand
Re:DeepSeek just open-sourced its inference engine
by hexa555 on Tuesday October 20, 2026 @07:53PM
(#50174308)
Isn't OpenRouter only part of the traffic of DeepSeek's inference?
Re:DeepSeek just open-sourced its inference engine
by brunoff on Tuesday October 20, 2026 @08:38PM
(#50174631)
Yes, it is only a part of DeepSeek's inference. Alongside openrouter there are others (probably many) doing this gateway work (for example opencode and even huggingface). Also the gateways in China itself, and of course also non gateway inference, DS itself directly providing inference.
Re:DeepSeek just open-sourced its inference engine
by richin13 on Tuesday October 20, 2026 @05:10PM
(#50172735)
OpenRouter routes traffic through many different providers
https://openrouter.ai/deepseek/deepseek-v4.1-flash-1017#prov...