Provider guide
Where to run Llama 3.1 70B Instruct
3 live listings tracked — output prices vary 1.8× between the cheapest and most expensive host, so the provider choice matters as much as the model choice.
FASTEST MEASURED
25 tok/s
measured throughput · $0.72 input / $0.72 output per million tokens
OpenRouterCHEAPEST$0.40 in/1M$0.40 out/1M—131K—5 hours agoAmazon Bedrockzero-retention$0.72 in/1M$0.72 out/1M25 tok/s131K—5 hours agoDeepInfraturbo tier$0.40 in/1M$0.40 out/1M17 tok/s131Kfp811 hours ago
DeepInfraFull specs, hardware verdicts and benchmarks on the Llama 3.1 70B Instruct model page →