Provider guide
Where to run Gemma 4 31B
16 live listings tracked — output prices vary 3.5× between the cheapest and most expensive host, so the provider choice matters as much as the model choice.
FASTEST MEASURED
126 tok/s
measured throughput · $0.75 input / $1.00 output per million tokens
DekaLLMzero-retentionCHEAPEST$0.10 in/1M$0.33 out/1M11 tok/s262K—14 hours agoOpenRouter$0.090 in/1M$0.34 out/1M—262K—2 hours agoCoreWeavezero-retention$0.10 in/1M$0.34 out/1M28 tok/s262Kfp42 hours agoVenice AIzero-retention$0.12 in/1M$0.36 out/1M18 tok/s256Kfp42 hours agoChutes$0.12 in/1M$0.37 out/1M19 tok/s131Kfp420 hours agoDeepInfrazero-retention through OpenRouterDirect and through OpenRouter$0.15 in/1M$0.40 out/1M10 tok/sthrough OpenRouter262Kfp82 hours agoNovita AIzero-retention through OpenRouterDirect and through OpenRouter$0.14 in/1M$0.40 out/1M8 tok/sthrough OpenRouter262Kbf162 hours ago direct4 days ago through OpenRouterCrusoezero-retention$0.14 in/1M$0.40 out/1M14 tok/s262Kbf162 hours agoParasailzero-retention$0.15 in/1M$0.40 out/1M17 tok/s262Kfp82 hours agoFriendli$0.14 in/1M$0.40 out/1M50 tok/s262K—2 hours agoSiliconFlowzero-retention$0.75 in/1M$1.00 out/1M38 tok/s262Kfp88 hours agoModelRunzero-retention$0.75 in/1M$1.00 out/1M126 tok/s262Kfp42 hours agoSambaNova$0.38 in/1M$1.15 out/1M—131K—2 hours agoIo Netzero-retention$0.38 in/1M$1.15 out/1M27 tok/s262K—2 hours agoSambaNovazero-retention$0.38 in/1M$1.15 out/1M28 tok/s262K—8 hours agoDeepInfraturbo tier$0.090 in/1M$0.34 out/1M23 tok/s262Kfp42 hours ago
OpenRouter
DeepInfra
Novita AI
SambaNovaFull specs, hardware verdicts and benchmarks on the Gemma 4 31B model page →