Skip to main content
4 DeepSeek text & code models on Kyma.

Which one

Default to deepseek-v4-flash for new work — it gives you 1M context, native reasoning, and the same family behaviour at a fraction of the cost. Move up to deepseek-v4-pro when quality matters more than price. Stay on deepseek-v3 if you have a production workload already tuned to it. deepseek-r1 is the reasoning specialist: slower, and worth it for maths and step-by-step analysis where you want the chain of thought exposed.