What is StreamLake
StreamLake provides hosted inference APIs for KAT-Coder and selected third-party language models, with OpenAI-compatible development workflows and usage-based billing.
Pay-as-you-go pricing
As of August 6, 2026, KAT-Coder-Pro-V2.5 costs US$0.74 per million input tokens, US$0.15 per million cached input tokens and US$2.96 per million output tokens. KAT-Coder-Air-V2.5 costs US$0.15 input, US$0.03 cached input and US$0.60 output, while KAT-Coder-Pro-V2 costs US$0.30 input, US$0.06 cached input and US$1.20 output.
Current third-party model rates per million tokens include DeepSeek V4 Flash at US$0.14 input, US$0.028 cached input and US$0.28 output; DeepSeek V4 Pro at US$1.74, US$0.145 and US$3.48; GLM 5.2 at US$1.40, US$0.26 and US$4.40; Qwen 3.5 35B at US$0.25, US$0.05 and US$2; and Kimi 2.6 at US$0.95, US$0.16 and US$4.
Coding Plan and billing
StreamLake also documents a quota-based KwaiKAT Coding Plan for supported development tools, but its public documentation does not expose a current numeric subscription price. The pricing page marks its older prepaid resource packs as offline, so they are not presented here as current purchase options. Buyers should compare the live model table and dashboard before committing to a workload.
Official sources: StreamLake pricing, StreamLake documentation.