Provider guide
Where to run Qwen3 30B A3B Instruct 2507
8 live listings tracked — output prices vary 2.7× between the cheapest and most expensive host, so the provider choice matters as much as the model choice.
FASTEST MEASURED
74 tok/s
measured throughput · $0.12 input / $0.50 output per million tokens
StreamLake$0.048 in/1M$0.19 out/1M68 tok/s128K—5 hours agoOpenRouterCHEAPEST$0.048 in/1M$0.19 out/1M—262K—5 hours agoDekaLLMzero-retention$0.090 in/1M$0.30 out/1M55 tok/s262K—5 hours agoSiliconFlowzero-retention$0.090 in/1M$0.30 out/1M20 tok/s262Kfp85 hours agoNebius AI Studiozero-retention$0.10 in/1M$0.30 out/1M59 tok/s262Kfp85 hours agoNovita AI$0.090 in/1M$0.45 out/1M—41K—21 days agoDeepInfrazero-retention through OpenRouterDirect and through OpenRouter$0.12 in/1M$0.50 out/1M74 tok/sthrough OpenRouter41Kfp85 hours agoAlibaba Cloud$0.13 in/1M$0.52 out/1M68 tok/s131K—5 hours ago
Nebius AI Studio
Novita AIFull specs, hardware verdicts and benchmarks on the Qwen3 30B A3B Instruct 2507 model page →