Provider guide
Where to run Qwen3 235B A22B Thinking 2507
8 live offers tracked — output prices vary 2.0× between the cheapest and most expensive host, so the provider choice matters as much as the model choice.
FASTEST MEASURED
69 tok/s
measured throughput · $0.23 input / $2.30 output per million tokens
Alibaba Cloud$0.15 in/1M$1.50 out/1M47 tok/s131K—7d agoDeepInfra$0.23 in/1M$2.30 out/1M—262Kfp87h agoOpenRouterCHEAPEST$0.23 in/1M$2.30 out/1M—262K—45m agoAlibaba Cloud$0.23 in/1M$2.30 out/1M69 tok/s131Kfp843m agoNovita AI$0.30 in/1M$3.00 out/1M—131K—7h agoDeepInfrazero-retention$0.23 in/1M$2.30 out/1M60 tok/s262Kfp843m agoNovita AIzero-retention$0.30 in/1M$3.00 out/1M33 tok/s131Kfp843m agoVenice AIzero-retention$0.45 in/1M$3.50 out/1M62 tok/s128Kfp843m ago
DeepInfra
Novita AIFull specs, hardware verdicts and benchmarks on theQwen3 235B A22B Thinking 2507 model page →