Provider guide
Where to run Qwen3 235B A22B Instruct 2507
15 live offers tracked — output prices vary 1.6× between the cheapest and most expensive host, so the provider choice matters as much as the model choice.
FASTEST MEASURED
43 tok/s
measured throughput · $0.15 input / $0.60 output per million tokens
DeepInfra$0.090 in/1M$0.55 out/1M—262Kfp87h agoNovita AI$0.090 in/1M$0.58 out/1M—131K—7h agoOpenRouterCHEAPEST$0.15 in/1M$0.60 out/1M—262K—45m agoAlibaba Cloud$0.15 in/1M$0.60 out/1M43 tok/s131Kfp843m agoAlibaba Cloud$0.15 in/1M$0.60 out/1M42 tok/s131K—7d agoFriendli$0.20 in/1M$0.80 out/1M32 tok/s262K—43m agoStreamLake$0.21 in/1M$0.84 out/1M31 tok/s128K—43m agoAtlasCloud$0.20 in/1M$0.88 out/1M33 tok/s131Kfp843m agoDeepInfrazero-retention$0.090 in/1M$0.55 out/1M18 tok/s262Kfp843m agoNovita AIzero-retention$0.090 in/1M$0.58 out/1M28 tok/s131Kfp843m agoNebius AI Studiozero-retention$0.20 in/1M$0.60 out/1M27 tok/s262Kfp88h agoVenice AIzero-retention$0.15 in/1M$0.75 out/1M16 tok/s128Kfp843m agoCrusoezero-retention$0.22 in/1M$0.80 out/1M34 tok/s262Kbf1643m agoParasailzero-retention$0.14 in/1M$0.80 out/1M28 tok/s131Kfp843m agoGoogle Vertex AIzero-retention$0.22 in/1M$0.88 out/1M29 tok/s262K—43m ago
DeepInfra
Novita AI
Nebius AI Studio
Google Vertex AIFull specs, hardware verdicts and benchmarks on theQwen3 235B A22B Instruct 2507 model page →