DeepSeek Fails to Deliver on Cheaper AI Claims Despite Nvidia Rental Test
The Call Center Doctors, a call center consultancy, recently tested the claims of DeepSeek's '80x cheaper' pricing. To do this, they rented four Nvidia H200s on September 27 to serve their Claude Code agents with DeepSeek V4.1 Flash instead of Opus 5.5.
The test aimed to see if DeepSeek could indeed be cheaper than the current solution. The consultancy found that at an on-demand cost, a whole box topped out at around 213 tokens written per second for $440.88 per day, versus $184-$223 per day for the same work via DeepSeek's API.
DeepSeek's API is priced at $0.15 per 1 million new input tokens off-peak, $0.003 cached input tokens, and $0.60 per 1 million output tokens. Peak costs are double. By comparison, Anthropic's Claude Opus 5.5 is listed at $4 input and $20 output per 1 million tokens, with cache reads at $0.20.
The consultancy used the DeepSeek solution through subscriptions and not the API. The test results showed that while the DeepSeek solution was cheaper in some cases, it wasn't as cheap as claimed, especially when considering security flaws kept the code offline.