DeepSeek API V4.1 Flash and V4 Pro pricing analyzed

An analysis of DeepSeek API pricing compares the V4.1 Flash and V4 Pro models against Claude and GPT coding workloads, incorporating peak and off-peak rates, token caching, and third-party provider costs. The author explains that V4.1 Flash is significantly cheaper than competing models even with zero cache hits, while V4 Pro becomes more cost-effective than GPT-6.1 Terra only when a substantial percentage of input tokens hit the cache during peak hours. Open article →