DeepSeek Made Crystal Clear · chapter 4: What DeepSeek Actually Costs

The pricing matrix

2026-10-23

The floor and ceiling for the same model and the same token type differ by roughly 63x, driven by two independent, controllable multipliers.

Below: the paragraph from the book that builds this idea, then the diagram itself (Figure 4.1), and a recap. About a minute of reading.

deepseek-v4-flash-vision-exp is priced identically to deepseek-v4-flash on every row. Look closely at the flash-to-pro ratio and you will find a small inconsistency worth stating rather than smoothing over: the cache-miss and output columns are exactly 3 times more expensive on v4-pro, but the cache-hit input columns are closer to 3.14 times, not a clean 3x. This book states the numbers as published rather than rounding them into a tidier pattern than they actually have.

Figure 4.1: The pricing matrix. The floor and ceiling for the same model and the same token type differ by roughly 63x, driven by two independent, controllable multipliers.
Figure 4.1: The pricing matrix. The floor and ceiling for the same model and the same token type differ by roughly 63x, driven by two independent, controllable multipliers.

Recap

  • The idea: The floor and ceiling for the same model and the same token type differ by roughly 63x, driven by two independent, controllable multipliers.
  • The picture: Figure 4.1, from chapter 4 ("What DeepSeek Actually Costs") of DeepSeek Made Crystal Clear.
  • Go deeper: the chapter builds this step by step, with recipes and sources at the end.

This diagram is one of many in DeepSeek Made Crystal Clear.

Every chapter opens with the gist, draws the hard ideas, and ends with recipes and sources.

Get the book

All diagrams