MixRoute says cached input for DeepSeek V4.1 Flash is now charged at $0.006 per million tokens, compared with $0.30 per million tokens for uncached input. According to the provider’s November 15 announcement on X, that makes cached input 50 times cheaper.
The figures apply specifically to input tokens. MixRoute’s post does not announce a corresponding change to output-token pricing, and the saving depends on requests actually qualifying for cached input. Customers with repeated prompts or shared context may therefore see a larger benefit than workloads dominated by unique input.
For API users, the distinction matters when comparing headline token prices: the uncached rate remains the relevant figure for ordinary input, while the cached rate applies only when the cache is used. Customers should verify the live model pricing and cache behaviour before estimating production costs.