Does Kilo Gateway support DeepSeek’s automatic Context Caching?
Hey everyone, I just ran a test comparing identical workloads on the exact same day directly through the DeepSeek API versus using the Kilo Gateway, and the price difference is shocking.
- Direct DeepSeek API: Cost $0.64 for the day. (DeepSeek’s automatic Context Caching kicked in, giving me nearly a 90% cache hit rate on my long prompts/code files).
- Kilo Gateway: Cost $2.34 for the exact same day.
This proves that Context Caching is completely missing or bypassed when routing queries through Kilo. You are paying full price for prompt inputs on every single request, which makes it nearly 4x more expensive for context-heavy tasks (like coding with large files open).
Recommendation: If you are doing heavy, context-repetitive tasks, stop using Kilo for now and switch to a direct DeepSeek API key. Otherwise, you are literally throwing money away on unsaved cache. Check your dashboards!
Does Kilo Gateway support DeepSeek’s automatic Context Caching, or does it bypass it?