DeepSeek

Now that deepseek-chat and deepseek-reasoner aliases are retired (July 24), you must explicitly call either deepseek-v4-flash ($0.14/$0.28/M, thinking ON by default) or deepseek-v4-pro ($0.44/$0.87/M, stronger reasoning). V4 Flash is 3x cheaper. V4 Pro is significantly stronger on complex tasks. Here is when each is the right choice.

DeepSeek V4 Pro ($0.44/$0.87/M discounted) and Gemini 3.6 Flash ($1.50/$7.50/M) are the two most cost-efficient capable production APIs in July 2026. V4 Pro is 3x cheaper on input. 3.6 Flash has Computer Use built in, US data residency, and 304 tok/s. V4 Pro defaults thinking ON and carries China data residency risk. Both have 1M context.