Provider guide
Where to run Qwen3.5 397B A17B
11 live listings tracked — output prices vary 1.9× between the cheapest and most expensive host, so the provider choice matters as much as the model choice.
FASTEST MEASURED
101 tok/s
measured throughput · $0.55 input / $3.50 output per million tokens
Alibaba CloudCHEAPEST$0.39 in/1M$2.34 out/1M73 tok/s262K—2 hours agoDeepInfrazero-retention through OpenRouterDirect and through OpenRouter$0.45 in/1M$3.00 out/1M17 tok/sthrough OpenRouter262Kfp88 hours ago direct2 hours ago through OpenRouterOpenRouter$0.55 in/1M$3.50 out/1M—262K—2 hours agoDigitalOcean Gradientzero-retention$0.55 in/1M$3.50 out/1M2 tok/s131K—8 hours agoAtlasCloud$0.55 in/1M$3.50 out/1M101 tok/s262Kfp82 hours agoPhalazero-retention$0.55 in/1M$3.50 out/1M34 tok/s262K—2 hours agoStreamLake$0.60 in/1M$3.60 out/1M73 tok/s256K—2 hours agoNovita AIzero-retention through OpenRouterDirect and through OpenRouter$0.60 in/1M$3.60 out/1M64 tok/sthrough OpenRouter262K—8 hours ago direct2 hours ago through OpenRouterGMICloud$0.60 in/1M$3.60 out/1M74 tok/s262Kfp88 hours agoParasailzero-retention$0.50 in/1M$3.60 out/1M24 tok/s262Kfp82 hours agoVenice AIzero-retention$0.75 in/1M$4.50 out/1M34 tok/s128K—2 hours ago
DeepInfra
OpenRouter
DigitalOcean Gradient
Novita AIFull specs, hardware verdicts and benchmarks on the Qwen3.5 397B A17B model page →