Provider guide
Where to run Qwen3.5 397B A17B
15 live offers tracked — output prices vary 1.5× between the cheapest and most expensive host, so the provider choice matters as much as the model choice.
FASTEST MEASURED
63 tok/s
measured throughput · $0.39 input / $2.34 output per million tokens
OpenRouterCHEAPEST$0.39 in/1M$2.34 out/1M—262K—18m agoAlibaba Cloud$0.39 in/1M$2.34 out/1M63 tok/s262K—6d agoAlibaba Cloud$0.39 in/1M$2.34 out/1M18 tok/s262Kfp816m agoDeepInfra$0.45 in/1M$3.00 out/1M—262Kfp86h agoChutes$0.45 in/1M$3.00 out/1M12 tok/s262Kfp816m agoAtlasCloud$0.55 in/1M$3.50 out/1M22 tok/s262Kfp816m agoNovita AI$0.60 in/1M$3.60 out/1M—262K—6h agoGMICloud$0.60 in/1M$3.60 out/1M—262Kfp816m agoStreamLake$0.60 in/1M$3.60 out/1M47 tok/s256K—16m agoDigitalOcean Gradientzero-retention$0.39 in/1M$2.45 out/1M9 tok/s131K—16m agoDeepInfrazero-retention$0.45 in/1M$3.00 out/1M11 tok/s262Kfp816m agoPhalazero-retention$0.55 in/1M$3.50 out/1M13 tok/s262K—16m agoNovita AIzero-retention$0.60 in/1M$3.60 out/1M56 tok/s262K—16m agoParasailzero-retention$0.50 in/1M$3.60 out/1M9 tok/s262Kfp816m agoVenice AIzero-retention$0.75 in/1M$4.50 out/1M57 tok/s128K—16m ago
DeepInfra
Novita AI
DigitalOcean GradientFull specs, hardware verdicts and benchmarks on theQwen3.5 397B A17B model page →