[discussion] DeepSeek via OpenRouter: same optimization performance as direct DeepSeek API?

[vlzsombor]

Hello Everyone,

Thanks for this project, I really enjoy using Whale.
I’m planning to use a DeepSeek model through OpenRouter (provider = "openai-compatible", base_url = "https://openrouter.ai/api/v1"), instead of calling DeepSeek directly.

I've read that 'Whale is DeepSeek-first — optimized for DeepSeek's caching, tools, and pricing'. But I am not sure if these optimization also applies if I use it through a model proxy.
Could you clarify if Whale’s caching behavior/performance will be the same as with direct DeepSeek?
I may have missed this in the docs — if it’s already documented, a link would be appreciated.

Thanks!