DeepSeek 開源推理引擎

Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

DeepSeek just open-sourced its inference engine (github.com/deepseek-ai) 106 points by ggcr1 hour ago | hide | past | favorite | 17 comments help wg010 minutes ago | next[–] That's Kami Kaze for US frontier labs. There's no way from here for them towards trillion dollar valuations. That's strategic sabotage to say the least. Ban all open weight models! Cried the VCs. reply dingaling9119 minutes ago | parent | next[–] That’s a bingo reply schneehertz46 minutes ago | prev | next[–] A nuclear bomb "Based on this framework, we have achieved high-throughput, low-latency large-scale inference services for DeepSeek V4.1 Flash. Currently, the service throughput of a single cluster is 155 trillion tokens/day, and the overall revenue is $38 per card per hour (at off-peak pricing)" Source: https://zhuanlan.zhihu.com/p/2095560295930508953 reply freakynit28 minutes ago | parent | next[–] For the ones who are still wondering how so much profits, inference is a highly profitable domain: https://inferencex.semianalysis.com/profit-estimator reply abound14 minutes ago | parent | prev | next[–] It's really refreshing to put some cold hard numbers on this stuff, hopefully this quells some of the "AI is unprofitable everyone is spending a dollar to make fifty cents" arguments. Amortized cost of owning a DC full of GPUs is far less than $38/card/hr Not to say it can't still be a bubble, just that there is legitimately profit to be had and perhaps some sustainable equilibrium in the future. reply EngineeringStuf8 minutes ago | parent | prev | next[–] $0.00051192 per M/tok ? Assuming a 64 card cluster reply jdw640 minutes ago | prev | next[–] I think the reason DeepSeek does this is partly to neutralize competitors' technology, but ultimately the core reason might be that data is what matters. reply aavaa5 minutes ago | prev | next[–] Notably missing their kv cache system reply ggcr1 hour ago | prev | next[–] https://x.com/lightseekorg/status/2112470406158561730 reply fukaiall15 minutes ago | prev | next[–] Not written in Rust? How dare you? reply pluc4 minutes ago | parent | next[–] Has to be something other than Rust to be able to be "rewritten in Rust" reply lukax6 minutes ago | parent | prev | next[–] Their inference server/service (wrapper around this engine) could still be written in Rust, this is just the engine. This is written in Rust: https://github.com/deepseek-ai/deepseek-recipe This DeepSeek Recipe is also mentioned here: https://zhuanlan.zhihu.com/p/2095560295930508953 reply cube0012 minutes ago | parent | prev | next[–] Given how much the models like React and JavaScript I guess we should be thankful its Python. reply baalimago14 minutes ago | prev[–] Apart from virtue signaling, why do this? I can't see how this benefits Deepseek reply dudisubekti12 minutes ago | parent[–] I'm curious, are projects like Linux "virtue signaling" to you? reply baalimago9 minutes ago | root | parent[–] No, and it's also not a major player in the AI race, backed by the Chinese state reply crimsoneer4 minutes ago | root | parent[–] It's releasing open source code. This is a good thing for the world, even if it impacts the portfolio of US VCs. reply Consider applying for YC's Winter 2027 batch! Applications are open till November 2. Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact Search: