It depends on the use-case. yes, 90% of cost is cache in agentic coding scenarios (actually 95% in my experience). But not when the model reasons for 200k+ tokens before answering a complex problem.

Gemini caching is confusing though:

  $0.15 / million tokens
  $1.00 / 1,000,000 tokens per hour (storage price)

In our experience, caching is not very reliable with google. We always get random cache misses that don't happen with other providers. We find OpenAI, Anthropic and Fireworks (which we use a lot) all have higher cache hit rates. So it's not only about the costs of cached token but also what kind of cached hit rate you get.

10% of input pricing is standard especially compared to competition.