What is StreamLake
StreamLake provides hosted inference APIs for multiple third-party language model families, with usage-based billing.
Pay-as-you-go pricing
The vendor's pricing table, updated September 17, 2026, lists rates per million tokens. DeepSeek V4 Flash is US$0.14 input, US$0.028 cached input and US$0.28 output; DeepSeek V4 Pro is US$1.74, US$0.145 and US$3.48; GLM 5.3 is US$1.40, US$0.26 and US$4.40; Qwen 3.8 2.4T is US$2 input, US$0.25 prefix-cached input and US$6 output; and Kimi K2.7 Code is US$0.95, US$0.19 and US$4. The full table also lists Qwen, MiniMax and MiMo variants with model-specific cache terms.
The current rate card no longer lists KAT-Coder model rates, so the previously displayed KAT-Coder prices have been removed. Pricing is model- and cache-type-specific; check the current vendor table before a large deployment.
Official sources: StreamLake pricing, StreamLake documentation.