Provider guide
Where to run DeepSeek V3.1 Terminus
13 live offers tracked — output prices vary 4.7× between the cheapest and most expensive host, so the provider choice matters as much as the model choice.
FASTEST MEASURED
58 tok/s
measured throughput · $0.34 input / $1.03 output per million tokens
AtlasCloud$0.30 in/1M$0.95 out/1M32 tok/s131Kfp85h agoOpenRouterCHEAPEST$0.25 in/1M$0.95 out/1M—164K—5h agoDeepInfra$0.25 in/1M$0.95 out/1M—164Kfp45h agoNovita AI$0.27 in/1M$1.00 out/1M—131K—5h agoStreamLake$0.34 in/1M$1.03 out/1M58 tok/s128K—5h agoSambaNova$3.00 in/1M$4.50 out/1M—131K—5h agoDeepInfrazero-retention$0.25 in/1M$0.95 out/1M8 tok/s164Kfp45h agoSiliconFlowzero-retention$0.27 in/1M$1.00 out/1M13 tok/s164Kfp85h agoNovita AIzero-retention$0.27 in/1M$1.00 out/1M26 tok/s131Kfp85h agoSambaNovazero-retention$0.65 in/1M$1.50 out/1M54 tok/s131Kfp85h agoCoreWeavezero-retention$0.55 in/1M$1.65 out/1M50 tok/s161Kfp85h agoGoogle Vertex AIzero-retention$0.60 in/1M$1.70 out/1M28 tok/s164K—5h agoMarazero-retention$0.60 in/1M$1.70 out/1M28 tok/s131K—43h ago
DeepInfra
Novita AI
SambaNova
Google Vertex AIFull specs, hardware verdicts and benchmarks on theDeepSeek V3.1 Terminus model page →