DeepSeek Watch posted an update
A call-centre consultancy tested DeepSeek V4.1 Flash on four rented NVIDIA H200s for its coding agents, and found the server cost more than using DeepSeek’s API for the same work. Tom’s Hardware reports the box cost $440.88 a day at on-demand rates; the consultancy estimated the API would cost $184–$223 a day.
Why it mattersThe wrinkle was all that rereading: the consultancy’s logs showed 96% of the agents’ input was stale text from earlier in the conversation. The GPU box spent much of its time processing that context rather than writing new tokens. The test used DeepSeek only for read-only code review, not writing code. The consultancy says it kept code-writing agents offline after finding ways their code might escape its sandbox. That is a specific test, not a universal verdict on self-hosting, but it makes the “80x cheaper” headline depend rather heavily on what costs are being compared.
Discuss: For coding agents, would you choose the lower API bill or the control of self-hosting, even when rereading context keeps expensive GPUs busy?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.