DeepSeek gave away the engine that runs its own chatbot.
DeepSeek open-sourced DeepInfer on Tuesday, the in-house software that serves its models in production. It runs on Nvidia's newest data center chips and on Huawei's Ascend 950, but supports only one model so far. Parts like caching and load balancing are missing, yet developers starred it over 9,000 times.
Why it mattersCompanies that host DeepSeek's models can now copy the tricks that keep its costs low. Official Huawei support also shows Chinese AI building its own software stack outside Nvidia's world.
Read in the app — free, in 9 languages
Related stories
MiniMax revealed that the mystery model Space Bunny was its own.2026-10-20Google's AI agents beat Minecraft on their own in 27 hours.
2026-10-19A breach at coding agent OpenCode exposed users' saved AI keys.
2026-10-18DeepSeek quietly made its cheap open model a far better coder.
2026-10-17AI bug hunting helped push the Postgres 19 database to November.
2026-10-17