Artificial Analysis Llama 3.3 Instruct 70B provider comparison (independent TTFT / tps / price measurements)

Independent third-party measurement of hosted-API latency (TTFT), output speed (tok/s), quality and price, continuously re-run across providers: for Llama 3.3 70B the benchmarked set includes Fireworks, Google Vertex, Groq, Novita, Parasail, SambaNova, CoreWeave, Together Turbo, Azure, DeepInfra (Turbo/FP8), Bedrock and Scaleway. The dynamic page did not yield per-cell numbers to static extraction on 2026-10-14; the OpenRouter provider table (svc-src-openrouter-provider-table-2026-10-14) supplies the corroborating measured P50s actually recorded, e.g. Groq ~233 tps median versus Groq's own 280 t/s marketing figure.